├── LICENSE ├── README.md ├── caffe-int8-convert-tool-dev.py ├── caffe-int8-convert-tool.py ├── classification-dev ├── googlenet_v1.table ├── mobilenet_v1.table ├── resnet18.table └── squeezenet_v1_1.table ├── classification ├── googlenet_v1.table ├── mobilenet_v1.table ├── mobilenet_v2.table ├── resnet18.table ├── resnet50.table └── squeezenet_v1_1.table └── detection └── mobilenet_v1-ssd-300.table /LICENSE: -------------------------------------------------------------------------------- 1 | BSD 3-Clause License 2 | 3 | Copyright (c) 2018, 4 | All rights reserved. 5 | 6 | Redistribution and use in source and binary forms, with or without 7 | modification, are permitted provided that the following conditions are met: 8 | 9 | * Redistributions of source code must retain the above copyright notice, this 10 | list of conditions and the following disclaimer. 11 | 12 | * Redistributions in binary form must reproduce the above copyright notice, 13 | this list of conditions and the following disclaimer in the documentation 14 | and/or other materials provided with the distribution. 15 | 16 | * Neither the name of the copyright holder nor the names of its 17 | contributors may be used to endorse or promote products derived from 18 | this software without specific prior written permission. 19 | 20 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 21 | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 23 | DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE 24 | FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 | DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 26 | SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 27 | CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 28 | OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 29 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Caffe-Int8-Convert-Tools 2 | 3 | This convert tools is base on TensorRT 2.0 Int8 calibration tools,which use the KL algorithm to find the suitable threshold to quantize the activions from Float32 to Int8(-128 - 127). 4 | 5 | We provide the Classification(SqueezeNet_v1.1) and Detection(MobileNet_v1 SSD 300) demos based on [ncnn](https://github.com/Tencent/ncnn)(It is a high-performance neural network inference framework optimized for the mobile platform),and the community ready to support this implment. 6 | 7 | [ncnn-int8](https://github.com/Tencent/ncnn/pull/487) 8 | 9 | ## Reference 10 | 11 | For details, please read the following PDF: 12 | 13 | [8-bit Inference with TensorRT](http://on-demand.gputechconf.com/gtc/2017/presentation/s7310-8-bit-inference-with-tensorrt.pdf) 14 | 15 | ## HowTo 16 | 17 | ### Release version 18 | 19 | ``` 20 | $ python caffe-int8-convert-tool.py --help 21 | usage: caffe-int8-convert-tool.py [-h] [--proto PROTO] [--model MODEL] 22 | [--mean MEAN MEAN MEAN] [--norm NORM] 23 | [--images IMAGES] [--output OUTPUT] 24 | [--gpu GPU] 25 | 26 | find the pretrained caffe models int8 quantize scale value 27 | 28 | optional arguments: 29 | -h, --help show this help message and exit 30 | --proto PROTO path to deploy prototxt. 31 | --model MODEL path to pretrained weights 32 | --mean MEAN value of mean 33 | --norm NORM value of normalize 34 | --images IMAGES path to calibration images 35 | --output OUTPUT path to output calibration table file 36 | --gpu GPU use gpu to forward 37 | 38 | $ python caffe-int8-convert-tool.py --proto=squeezenet_v1.1.prototxt --model=squeezenet.caffemodel --mean 104 117 123 --images=ILSVRC2012_1k --output=squeezenet_v1.1.table --gpu=1 39 | ``` 40 | 41 | Pay attention to the type of images,it is just the original image format,such as jpg or jpeg,do not us the type of caffe dataset(lmdb). 42 | **It is recommended to provide representative calibration dataset for the given model use case,such as the validation or test dataset.** 43 | 44 | ### How to use the output file(calibration.table) 45 | 46 | For example in *squeezenet_v1_1.table* 47 | 48 | ``` 49 | conv1_param_0 138.066410 50 | fire2/squeeze1x1_param_0 92.028103 // the conv layer's weight scale is 92.028 51 | ...... 52 | data 0.841264 53 | conv1 0.295743 54 | pool1 0.161700 // the pool layer's top blob scale is 0.1617 55 | fire2/squeeze1x1 0.089383 // the conv layer's top blob scale is 0.0893 56 | ...... 57 | ``` 58 | 59 | Three steps to implement the *fire2/squeeze1x1* layer int8 convolution: 60 | 61 | 1. Quantize the bottom_blob and weight: 62 | 63 | ``` 64 | bottom_blob_int8 = bottom_blob_float32 * data_scale(0.1617) 65 | weight_int8 = weight_float32 * weight_scale(92.028) 66 | ``` 67 | 68 | 2. Convolution_Int8: 69 | 70 | ``` 71 | top_blob_int32 = bottom_blob_int8 * weight_int8 72 | ``` 73 | 74 | 3. Dequantize the TopBlob_Int32 and add the bias: 75 | 76 | ``` 77 | top_blob_float32 = top_blob_int32 / [data_scale(0.1617) * weight_scale(92.028)] + bias_float32 78 | ``` 79 | 80 | ### Development version 81 | 82 | The purpose of this tool(caffe-int8-convert-tool-dev.py) is to test new features,such as mulit-channels quantization depend on group num,sparse calculation and so on. 83 | 84 | This format is already supported in the [ncnn](https://github.com/Tencent/ncnn) latest version.I will do my best to transform some common network models into [classification-dev](https://github.com/BUG1989/caffe-int8-convert-tools/tree/master/classification-dev) 85 | 86 | ``` 87 | python caffe-int8-convert-tool-dev.py -h 88 | usage: caffe-int8-convert-tool.py [-h] [--proto PROTO] [--model MODEL] 89 | [--mean MEAN MEAN MEAN] [--norm NORM] 90 | [--images IMAGES] [--output OUTPUT] 91 | [--group GROUP] [--gpu GPU] 92 | 93 | find the pretrained caffe models int8 quantize scale value 94 | 95 | optional arguments: 96 | -h, --help show this help message and exit 97 | --proto PROTO path to deploy prototxt. 98 | --model MODEL path to pretrained weights 99 | --mean MEAN value of mean 100 | --norm NORM value of normalize 101 | --images IMAGES path to calibration images 102 | --output OUTPUT path to output calibration table file 103 | --group GROUP enable the group scale 104 | --gpu GPU use gpu to forward 105 | python caffe-int8-convert-tool-dev.py --proto=test/models/mobilenet_v1.prototxt --model=test/models/mobilenet_v1.caffemodel --mean 103.94 116.78 123.68 --norm=0.017 --images=test/images/ --group=1 106 | ``` 107 | 108 | Although it's done,but the speed of group quanization is very slow......The difference from the release tool is that we try to get the int8_scale of bottom blob not the top blob. 109 | 110 | ### How to use the output file(calibration-dev.table) 111 | 112 | For example in *MobileNet_v1_dev.table* 113 | 114 | ``` 115 | conv1_param_0 156.639840 116 | conv2_1/dw_param_0 0 72.129143 149.919382 // the convdw layer's weight scale every group is 0.0 72.129 149.919 ...... 117 | ...... 118 | conv1 49.466518 119 | conv2_1/dw 0 123.720796 48.705349 ...... // the convdw layer's bottom blob every group channel scale is 0.0 123.720 48.705 ...... 120 | ...... 121 | ``` 122 | 123 | ## Accuracy and Performance 124 | 125 | We used ImageNet2012 Dataset to complete some experiments. 126 | 127 | | Type | Detail | 128 | | ------------------- | ----------------------------------------------------- | 129 | | Calibration Dataset | ILSVRC2012_img_test   1k | 130 | | Test Dataset | ILSVRC2012_img_val     5k | 131 | | Framework | ncnn-int8 | 132 | | Support Layer | Convolution3x3,Convolution1x1,ConvolutionDepthwise3x3 | 133 | 134 | The following table show the Top1 and Top5 different between Float32 and Int8 inference. 135 | 136 | | | FP32 | | INT8 | | 137 | | --------------- | ------ | ------ | --------- | --------- | 138 | | NETWORK | Top1 | Top5 | Top1 | Top5 | 139 | | SqueezeNet v1.1 | 57.86% | 79.86% | 57.36% | 79.84% | 140 | | MobileNet v1 | 67.78% | 87.62% | 64.92% | 85.22% | 141 | | MobileNet v2 | 70.20% | 89.20% | 69.00% | 88.04% | 142 | | GoogleNet v1 | 67.70% | 88.32% | 67.64% | 88.26% | 143 | | ResNet-18 | 65.50% | 86.46% | 65.48% | 86.44% | 144 | | ResNet-50 | 71.68% | 89.94% | 71.38% | 89.52% | 145 | | NETWORK | Top1 | Top5 | Diff Top1 | Diff Top5 | 146 | | SqueezeNet v1.1 | 57.86% | 79.86% | 0.50% | 0.02% | 147 | | MobileNet v1 | 67.78% | 87.62% | 2.86% | 2.40% | 148 | | MobileNet v2 | 70.20% | 89.20% | 1.06% | 1.16% | 149 | | GoogleNet v1 | 67.70% | 88.32% | 0.06% | 0.06% | 150 | | ResNet-18 | 65.50% | 86.46% | 0.02% | 0.02% | 151 | | ResNet-50 | 71.68% | 89.94% | 0.30% | 0.32% | 152 | 153 | The following table show the speedup between Float32 and Int8 inference.It should be noted that the winograd algorithm is not used in the Float32 inference.The Hardware Platform is Hisi3519(Cortex-A17@1.2GHz) 154 | 155 | | Uint(ms) | SqueezeNet v1.1 | MobileNet v1 | MobileNet v2 | GoogleNet | ResNet18 | MobileNetv1 SSD | 156 | | -------- | --------------- | ------------ | ------------ | --------- | -------- | --------------- | 157 | | Float32 | 382 | 568 | 392 | 1662 | 1869 | 1120 | 158 | | Int8 | 242 | 369 | 311 | 1159 | 1159 | 701 | 159 | | Ratio | x1.30 | x1.41 | x1.28 | x1.43 | x1.61 | x1.47 | 160 | 161 | ## Contributor 162 | 163 | Thanks to our company [SenseNets](http://www.sensenets.com/home/) to support the open source project,and NVIDIA for providing the principle of correlation entropy,and ncnn's author [nihui](https://github.com/nihui) sharing his neural network inference framework. 164 | 165 | Thanks to the help from the following friends: 166 | 167 | Algorithm : [xupengfeixupf](https://github.com/xupengfeixupf), [JansonZhu](https://github.com/JansonZhu), wangxinwei, [lengmm](https://github.com/lengmm) 168 | 169 | Python : [daquexian](https://github.com/daquexian) 170 | 171 | ## License 172 | 173 | BSD 3 Clause 174 | 175 | -------------------------------------------------------------------------------- /caffe-int8-convert-tool-dev.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | 3 | # SenseNets is pleased to support the open source community by making caffe-int8-convert-tool available. 4 | # 5 | # Copyright (C) 2018 SenseNets Technology Ltd. All rights reserved. 6 | # 7 | # Licensed under the BSD 3-Clause License (the "License"); you may not use this file except 8 | # in compliance with the License. You may obtain a copy of the License at 9 | # 10 | # https://opensource.org/licenses/BSD-3-Clause 11 | # 12 | # Unless required by applicable law or agreed to in writing, software distributed 13 | # under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR 14 | # CONDITIONS OF ANY KIND, either express or implied. See the License for the 15 | # specific language governing permissions and limitations under the License. 16 | 17 | 18 | """ 19 | Quantization module for generating the calibration tables will be used by 20 | quantized (INT8) models from FP32 models. 21 | This tool is based on Caffe Framework. 22 | """ 23 | from __future__ import division 24 | from __future__ import print_function 25 | import argparse 26 | import numpy as np 27 | import math, copy 28 | import matplotlib.pyplot as plt 29 | import sys,os 30 | import caffe 31 | import caffe.proto.caffe_pb2 as caffe_pb2 32 | import time 33 | import datetime 34 | from google.protobuf import text_format 35 | 36 | 37 | def parse_args(): 38 | parser = argparse.ArgumentParser( 39 | description='find the pretrained caffe models int8 quantize scale value') 40 | parser.add_argument('--proto', dest='proto', 41 | help="path to deploy prototxt.", type=str) 42 | parser.add_argument('--model', dest='model', 43 | help='path to pretrained weights', type=str) 44 | parser.add_argument('--mean', dest='mean', 45 | help='value of mean', type=float, nargs=3) 46 | parser.add_argument('--norm', dest='norm', 47 | help='value of normalize', type=float, nargs=1, default=1.0) 48 | parser.add_argument('--images', dest='images', 49 | help='path to calibration images', type=str) 50 | parser.add_argument('--output', dest='output', 51 | help='path to output calibration table file', type=str, default='calibration-dev.table') 52 | parser.add_argument('--group', dest='group', 53 | help='enable the group scale', type=int, default=0) 54 | parser.add_argument('--gpu', dest='gpu', 55 | help='use gpu to forward', type=int, default=0) 56 | 57 | args = parser.parse_args() 58 | return args, parser 59 | 60 | 61 | global args, parser 62 | args, parser = parse_args() 63 | 64 | 65 | # global params 66 | QUANTIZE_NUM = 127 67 | STATISTIC = 1 68 | INTERVAL_NUM = 2048 69 | 70 | # ugly global params 71 | quantize_layer_lists = [] 72 | 73 | 74 | class QuantizeLayer: 75 | def __init__(self, name, blob_name, group_num): 76 | self.name = name 77 | self.blob_name = blob_name 78 | self.group_num = group_num 79 | self.weight_scale = [0 for x in range(0, group_num)] 80 | self.blob_max = [0 for x in range(0, group_num)] 81 | self.blob_distubution_interval = [0 for x in range(0, group_num)] 82 | self.blob_distubution = [[0 for col in range(INTERVAL_NUM)] for row in range(group_num)] 83 | self.blob_scale = [1 for x in range(0, group_num)] 84 | self.group_zero = [0 for x in range(0, group_num)] 85 | 86 | def quantize_weight(self, weight_data): 87 | # spilt the weight data by group num 88 | blob_group_data = np.array_split(weight_data, self.group_num) 89 | for i, group_data in enumerate(blob_group_data): 90 | max_val = np.max(group_data) 91 | min_val = np.min(group_data) 92 | threshold = max(abs(max_val), abs(min_val)) 93 | if threshold < 0.0001: 94 | self.weight_scale[i] = 0 95 | self.group_zero[i] = 1 96 | else: 97 | self.weight_scale[i] = QUANTIZE_NUM / threshold 98 | print("%-20s group : %-5d max_val : %-10f scale_val : %-10f" % (self.name + "_param0", i, threshold, self.weight_scale[i])) 99 | 100 | def initial_blob_max(self, blob_data): 101 | # spilt the blob data by group num 102 | blob_group_data = np.array_split(blob_data, self.group_num) 103 | # interval for per bottom blob group channel 104 | for i, group_data in enumerate(blob_group_data): 105 | max_val = np.max(group_data) 106 | min_val = np.min(group_data) 107 | self.blob_max[i] = max(self.blob_max[i], max(abs(max_val), abs(min_val))) 108 | 109 | def initial_blob_distubution_interval(self): 110 | for i in range(0, self.group_num): 111 | if self.blob_max[i] < 0.000001: 112 | self.blob_scale[i] = 0 113 | self.group_zero[i] = 1 114 | self.blob_distubution_interval[i] = 0 115 | else: 116 | self.blob_distubution_interval[i] = STATISTIC * self.blob_max[i] / INTERVAL_NUM 117 | print("%-20s group : %-5d max_val : %-10.8f distribution_intervals : %-10.8f" % (self.name, i, self.blob_max[i], self.blob_distubution_interval[i])) 118 | 119 | def initial_histograms(self, blob_data): 120 | # spilt the blob data by group num 121 | blob_group_data = np.array_split(blob_data, self.group_num) 122 | # interval for per bottom blob group channel 123 | for i, group_data in enumerate(blob_group_data): 124 | if self.blob_scale[i] == 0: 125 | continue 126 | else: 127 | # collect histogram of every group channel blob 128 | add_to_distribution(group_data, self.blob_distubution[i], self.blob_distubution_interval[i]) 129 | 130 | def quantize_blob(self): 131 | # calculate threshold 132 | for i in range(0, self.group_num): 133 | # sparse DepthwiseConvolution 134 | if self.blob_scale[i] == 0: 135 | print("%-20s group : %-5d bin : %-8d threshold : %-10f interval : %-10f scale : %-10f" % (self.name, i, 0, 0, self.blob_distubution_interval[i], self.blob_scale[i])) 136 | else: 137 | # normalize distributions 138 | normalize_distribution(self.blob_distubution[i]) 139 | distribution = np.array(self.blob_distubution[i]) 140 | # pick threshold which minimizes KL divergence 141 | threshold_bin = threshold_distribution(distribution) 142 | threshold = (threshold_bin + 0.5) * self.blob_distubution_interval[i] 143 | # get the activation calibration value 144 | self.blob_scale[i] = QUANTIZE_NUM / threshold 145 | print("%-20s group : %-5d bin : %-8d threshold : %-10f interval : %-10f scale : %-10f" % (self.name, i, threshold_bin, threshold, self.blob_distubution_interval[i], self.blob_scale[i])) 146 | 147 | def display_sparse_info(self): 148 | count = 0 149 | for i in range(self.group_num): 150 | if self.group_zero[i] != 0: 151 | count += 1 152 | print("%-20s group total : %-8d group sparse : %-8d ratio : %-6.2f " % (self.name, self.group_num, count, count / float(self.group_num) * 100)) 153 | 154 | def save_calibration(file_path): 155 | pass 156 | 157 | 158 | 159 | def add_to_distribution(blob, distribution, interval): 160 | """ 161 | add the distribution 162 | Args: 163 | blob: the output blob of caffe layer 164 | distribution: a list ,size is 2048 165 | interval: a float number 166 | Returns: 167 | none 168 | """ 169 | max_index = len(distribution) - 1 170 | 171 | indexes = np.minimum((np.abs(blob[blob!=0]) / interval).astype(np.int32), max_index) 172 | for index in indexes: 173 | distribution[index] = distribution[index] + 1 174 | 175 | 176 | def normalize_distribution(distribution): 177 | """ 178 | Normalize the input list 179 | Args: 180 | distribution: a list ,size is 2048 181 | Returns: 182 | none 183 | """ 184 | num_sum = sum(distribution) 185 | for i, data in enumerate(distribution): 186 | distribution[i] = data / float(num_sum) 187 | 188 | 189 | def compute_kl_divergence(dist_a, dist_b): 190 | """ 191 | Returen kl_divergence between 192 | Args: 193 | dist_a: list original 194 | dist_b: list expand 195 | Returns: 196 | kl_divergence: float, kl_divergence 197 | """ 198 | nonzero_inds = dist_a != 0 199 | return np.sum(dist_a[nonzero_inds] * np.log(dist_a[nonzero_inds] / dist_b[nonzero_inds])) 200 | 201 | 202 | def threshold_distribution(distribution, target_bin=128): 203 | """ 204 | Returen the best cut off num of bin 205 | Args: 206 | distribution: list, activations has been processed by histogram and normalize,size is 2048 207 | target_bin: int, the num of bin that is used by quantize, Int8 default value is 128 208 | Returns: 209 | target_threshold: int, num of bin with the minimum KL 210 | """ 211 | target_threshold = target_bin 212 | min_kl_divergence = 1000 213 | length = distribution.size 214 | 215 | quantize_distribution = np.zeros(target_bin) 216 | 217 | threshold_sum = 0.0 218 | threshold_sum = sum(distribution[target_bin:]) 219 | 220 | for threshold in range(target_bin, length): 221 | t_distribution = copy.deepcopy(distribution[:threshold]) 222 | t_distribution[threshold-1] = t_distribution[threshold-1] + threshold_sum 223 | threshold_sum = threshold_sum - distribution[threshold] 224 | 225 | # ************************ threshold ************************ 226 | quantize_distribution = np.zeros(target_bin) 227 | num_per_bin = threshold / target_bin 228 | for i in range(0, target_bin): 229 | start = i * num_per_bin 230 | end = start + num_per_bin 231 | 232 | left_upper = (int)(math.ceil(start)) 233 | if(left_upper > start): 234 | left_scale = left_upper - start 235 | quantize_distribution[i] += left_scale * distribution[left_upper - 1] 236 | 237 | right_lower = (int)(math.floor(end)) 238 | if (right_lower < end): 239 | right_scale = end - right_lower 240 | quantize_distribution[i] += right_scale * distribution[right_lower] 241 | 242 | for j in range(left_upper, right_lower): 243 | quantize_distribution[i] += distribution[j] 244 | # ************************ threshold ************************ 245 | 246 | # ************************ quantzie ************************ 247 | expand_distribution = np.zeros(threshold, dtype=np.float32) 248 | 249 | for i in range(0, target_bin): 250 | start = i * num_per_bin 251 | end = start + num_per_bin 252 | 253 | count = 0 254 | 255 | left_upper = (int)(math.ceil(start)) 256 | left_scale = 0.0 257 | if (left_upper > start): 258 | left_scale = left_upper - start 259 | if (distribution[left_upper - 1] != 0): 260 | count += left_scale 261 | 262 | right_lower = (int)(math.floor(end)) 263 | right_scale = 0.0 264 | if (right_lower < end): 265 | right_scale = end - right_lower 266 | if (distribution[right_lower] != 0): 267 | count += right_scale 268 | 269 | for j in range(left_upper, right_lower): 270 | if (distribution[j] != 0): 271 | count = count + 1 272 | 273 | if count == 0: 274 | continue; 275 | expand_value = quantize_distribution[i] / count 276 | 277 | if (left_upper > start): 278 | if (distribution[left_upper - 1] != 0): 279 | expand_distribution[left_upper - 1] += expand_value * left_scale 280 | if (right_lower < end): 281 | if (distribution[right_lower] != 0): 282 | expand_distribution[right_lower] += expand_value * right_scale 283 | for j in range(left_upper, right_lower): 284 | if (distribution[j] != 0): 285 | expand_distribution[j] += expand_value 286 | # ************************ quantzie ************************ 287 | 288 | kl_divergence = compute_kl_divergence(t_distribution, expand_distribution) 289 | 290 | if kl_divergence < min_kl_divergence: 291 | min_kl_divergence = kl_divergence 292 | target_threshold = threshold 293 | 294 | return target_threshold 295 | 296 | 297 | def net_forward(net, image_path, transformer): 298 | """ 299 | network inference and statistics the cost time 300 | Args: 301 | net: the instance of Caffe inference 302 | image_path: a image need to be inference 303 | transformer: 304 | Returns: 305 | none 306 | """ 307 | # load image 308 | image = caffe.io.load_image(image_path) 309 | # transformer.preprocess the image 310 | net.blobs['data'].data[...] = transformer.preprocess('data',image) 311 | # net forward 312 | start = time.clock() 313 | output = net.forward() 314 | end = time.clock() 315 | print("%s forward time : %.3f s" % (image_path, end - start)) 316 | 317 | 318 | def file_name(file_dir): 319 | """ 320 | Find the all file path with the directory 321 | Args: 322 | file_dir: The source file directory 323 | Returns: 324 | files_path: all the file path into a list 325 | """ 326 | files_path = [] 327 | 328 | for root, dir, files in os.walk(file_dir): 329 | for name in files: 330 | file_path = root + "/" + name 331 | print(file_path) 332 | files_path.append(file_path) 333 | 334 | return files_path 335 | 336 | 337 | def network_prepare(net, mean, norm): 338 | """ 339 | instance the prepare process param of caffe network inference 340 | Args: 341 | net: the instance of Caffe inference 342 | mean: the value of mean 343 | norm: the value of normalize 344 | Returns: 345 | none 346 | """ 347 | print("Network initial") 348 | 349 | img_mean = np.array(mean) 350 | 351 | # initial transformer 352 | transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape}) 353 | # convert shape from RBG to BGR 354 | transformer.set_transpose('data', (2,0,1)) 355 | # load meanfile 356 | transformer.set_mean('data', img_mean) 357 | # resize image data from [0,1] to [0,255] 358 | transformer.set_raw_scale('data', 255) 359 | # convert RGB -> BGR 360 | transformer.set_channel_swap('data', (2,1,0)) 361 | # normalize 362 | transformer.set_input_scale('data', norm) 363 | 364 | return transformer 365 | 366 | 367 | def weight_quantize(net, net_file, group_on): 368 | """ 369 | CaffeModel convolution weight blob Int8 quantize 370 | Args: 371 | net: the instance of Caffe inference 372 | net_file: deploy caffe prototxt 373 | Returns: 374 | none 375 | """ 376 | print("\nQuantize the kernel weight:") 377 | 378 | # parse the net param from deploy prototxt 379 | params = caffe_pb2.NetParameter() 380 | with open(net_file) as f: 381 | text_format.Merge(f.read(), params) 382 | 383 | for i, layer in enumerate(params.layer): 384 | if i == 0: 385 | if layer.type != "Input": 386 | raise ValueError("First layer should be input") 387 | 388 | # find the convolution 3x3 and 1x1 layers to get out the weight_scale 389 | if(layer.type == "Convolution" or layer.type == "ConvolutionDepthwise"): 390 | kernel_size = layer.convolution_param.kernel_size[0] 391 | if(kernel_size == 3 or kernel_size == 1): 392 | weight_blob = net.params[layer.name][0].data 393 | # initial the instance of QuantizeLayer Class lists,you can use enable group quantize to generate int8 scale for each group layer.convolution_param.group 394 | if (group_on == 1): 395 | quanitze_layer = QuantizeLayer(layer.name, layer.bottom[0], layer.convolution_param.group) 396 | else: 397 | quanitze_layer = QuantizeLayer(layer.name, layer.bottom[0], 1) 398 | # quantize the weight value 399 | quanitze_layer.quantize_weight(weight_blob) 400 | # add the quantize_layer into the save list 401 | quantize_layer_lists.append(quanitze_layer) 402 | 403 | return None 404 | 405 | 406 | def activation_sparse(net, transformer, images_files): 407 | """ 408 | Activation bottom blob sparse analyze 409 | Args: 410 | net: the instance of Caffe inference 411 | transformer: 412 | images_files: calibration dataset 413 | Returns: 414 | none 415 | """ 416 | print("\nAnalyze the sparse info of the Activation:") 417 | # run float32 inference on calibration dataset to find the activations range 418 | for i , image in enumerate(images_files): 419 | net_forward(net, image, transformer) 420 | print("loop stage 1 : %d" % (i)) 421 | # find max threshold 422 | for layer in quantize_layer_lists: 423 | blob = net.blobs[layer.blob_name].data[0].flatten() 424 | layer.initial_blob_max(blob) 425 | 426 | # calculate statistic blob scope and interval distribution 427 | for layer in quantize_layer_lists: 428 | layer.initial_blob_distubution_interval() 429 | 430 | return None 431 | 432 | 433 | def activation_quantize(net, transformer, images_files): 434 | """ 435 | Activation Int8 quantize, optimaize threshold selection with KL divergence, 436 | given a dataset, find the optimal threshold for quantizing it. 437 | Ref: http://on-demand.gputechconf.com/gtc/2017/presentation/s7310-8-bit-inference-with-tensorrt.pdf 438 | Args: 439 | net: the instance of Caffe inference 440 | transformer: 441 | images_files: calibration dataset 442 | Returns: 443 | none 444 | """ 445 | print("\nQuantize the Activation:") 446 | # run float32 inference on calibration dataset to find the activations range 447 | for i , image in enumerate(images_files): 448 | net_forward(net, image, transformer) 449 | print("loop stage 1 : %d" % (i)) 450 | # find max threshold 451 | for layer in quantize_layer_lists: 452 | blob = net.blobs[layer.blob_name].data[0].flatten() 453 | layer.initial_blob_max(blob) 454 | 455 | # calculate statistic blob scope and interval distribution 456 | for layer in quantize_layer_lists: 457 | layer.initial_blob_distubution_interval() 458 | 459 | # for each layers 460 | # collect histograms of activations 461 | print("\nCollect histograms of activations:") 462 | for i, image in enumerate(images_files): 463 | net_forward(net, image, transformer) 464 | print("loop stage 2 : %d" % (i)) 465 | start = time.clock() 466 | for layer in quantize_layer_lists: 467 | blob = net.blobs[layer.blob_name].data[0].flatten() 468 | layer.initial_histograms(blob) 469 | end = time.clock() 470 | print("add cost %.3f s" % (end - start)) 471 | 472 | # calculate threshold with KL divergence 473 | for layer in quantize_layer_lists: 474 | layer.quantize_blob() 475 | 476 | return None 477 | 478 | 479 | def save_calibration_file(calibration_path): 480 | calibration_file = open(calibration_path, 'w') 481 | # save temp 482 | save_temp = [] 483 | # save weight scale 484 | for layer in quantize_layer_lists: 485 | save_string = layer.name + "_param_0" 486 | for i in range(layer.group_num): 487 | save_string = save_string + " " + str(layer.weight_scale[i]) 488 | save_temp.append(save_string) 489 | 490 | # save bottom blob scales 491 | for layer in quantize_layer_lists: 492 | save_string = layer.name 493 | for i in range(layer.group_num): 494 | save_string = save_string + " " + str(layer.blob_scale[i]) 495 | save_temp.append(save_string) 496 | 497 | # save into txt file 498 | for data in save_temp: 499 | calibration_file.write(data + "\n") 500 | 501 | calibration_file.close() 502 | 503 | 504 | def usage_info(): 505 | """ 506 | usage info 507 | """ 508 | print("Input params is illegal...╮(╯3╰)╭") 509 | print("try it again:\n python caffe-int8-scale-tools.py -h") 510 | 511 | 512 | def main(): 513 | """ 514 | main function 515 | """ 516 | 517 | # time start 518 | time_start = datetime.datetime.now() 519 | 520 | print(args) 521 | 522 | if args.proto == None or args.model == None or args.mean == None or args.images == None: 523 | usage_info() 524 | return None 525 | 526 | # deploy caffe prototxt path 527 | net_file = args.proto 528 | 529 | # trained caffemodel path 530 | caffe_model = args.model 531 | 532 | # mean value 533 | mean = args.mean 534 | 535 | # norm value 536 | norm = 1.0 537 | if args.norm != 1.0: 538 | norm = args.norm[0] 539 | 540 | # calibration dataset 541 | images_path = args.images 542 | 543 | # the output calibration file 544 | calibration_path = args.output 545 | 546 | # enable the group scale 547 | group_on = args.group 548 | 549 | # default use CPU to forwark 550 | if args.gpu != 0: 551 | caffe.set_device(0) 552 | caffe.set_mode_gpu() 553 | 554 | # initial caffe net and the forword model(GPU or CPU) 555 | net = caffe.Net(net_file,caffe_model,caffe.TEST) 556 | 557 | # prepare the cnn network 558 | transformer = network_prepare(net, mean, norm) 559 | 560 | # get the calibration datasets images files path 561 | images_files = file_name(images_path) 562 | 563 | # quanitze kernel weight of the caffemodel to find it's calibration table 564 | # weight_quantize(net, net_file) 565 | weight_quantize(net, net_file, group_on) 566 | 567 | # quantize activation value of the caffemodel to find it's calibration table 568 | activation_quantize(net, transformer, images_files) 569 | 570 | # save the calibration tables,best wish for your INT8 inference have low accuracy loss :) 571 | save_calibration_file(calibration_path) 572 | 573 | # time end 574 | time_end = datetime.datetime.now() 575 | 576 | print("\nCaffe Int8 Calibration table create success, it's cost %s, best wish for your INT8 inference has a low accuracy loss...\(^▽^)/...2333..." % (time_end - time_start)) 577 | 578 | if __name__ == "__main__": 579 | main() 580 | -------------------------------------------------------------------------------- /caffe-int8-convert-tool.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | 3 | # SenseNets is pleased to support the open source community by making caffe-int8-convert-tool available. 4 | # 5 | # Copyright (C) 2018 SenseNets Technology Ltd. All rights reserved. 6 | # 7 | # Licensed under the BSD 3-Clause License (the "License"); you may not use this file except 8 | # in compliance with the License. You may obtain a copy of the License at 9 | # 10 | # https://opensource.org/licenses/BSD-3-Clause 11 | # 12 | # Unless required by applicable law or agreed to in writing, software distributed 13 | # under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR 14 | # CONDITIONS OF ANY KIND, either express or implied. See the License for the 15 | # specific language governing permissions and limitations under the License. 16 | 17 | 18 | """ 19 | Quantization module for generating the calibration tables will be used by 20 | quantized (INT8) models from FP32 models. 21 | This tool is based on Caffe Framework. 22 | """ 23 | from __future__ import division 24 | from __future__ import print_function 25 | import argparse 26 | import numpy as np 27 | import math, copy 28 | import matplotlib.pyplot as plt 29 | import sys,os 30 | import caffe 31 | import caffe.proto.caffe_pb2 as caffe_pb2 32 | import time 33 | from google.protobuf import text_format 34 | 35 | 36 | def parse_args(): 37 | parser = argparse.ArgumentParser( 38 | description='find the pretrained caffe models int8 quantize scale value') 39 | parser.add_argument('--proto', dest='proto', 40 | help="path to deploy prototxt.", type=str) 41 | parser.add_argument('--model', dest='model', 42 | help='path to pretrained weights', type=str) 43 | parser.add_argument('--mean', dest='mean', 44 | help='value of mean', type=float, nargs=3) 45 | parser.add_argument('--norm', dest='norm', 46 | help='value of normalize', type=float, nargs=1, default=1.0) 47 | parser.add_argument('--images', dest='images', 48 | help='path to calibration images', type=str) 49 | parser.add_argument('--output', dest='output', 50 | help='path to output calibration table file', type=str, default='calibration.table') 51 | parser.add_argument('--gpu', dest='gpu', 52 | help='use gpu to forward', type=int, default=0) 53 | 54 | args = parser.parse_args() 55 | return args, parser 56 | 57 | 58 | global args, parser 59 | args, parser = parse_args() 60 | 61 | 62 | # global params 63 | QUANTIZE_NUM = 127 64 | STATISTIC = 1 65 | INTERVAL_NUM = 2048 66 | 67 | # ugly global params 68 | top_blob_names = [] 69 | distribution_intervals = [] 70 | save_temp = [] 71 | 72 | 73 | def add_to_distribution(blob, distribution, interval): 74 | """ 75 | add the distribution 76 | Args: 77 | blob: the output blob of caffe layer 78 | distribution: a list ,size is 2048 79 | interval: a float number 80 | Returns: 81 | none 82 | """ 83 | max_index = len(distribution) - 1 84 | 85 | indexes = np.minimum((np.abs(blob[blob!=0]) / interval).astype(np.int32), max_index) 86 | for index in indexes: 87 | distribution[index] = distribution[index] + 1 88 | 89 | 90 | def normalize_distribution(distribution): 91 | """ 92 | Normalize the input list 93 | Args: 94 | distribution: a list ,size is 2048 95 | Returns: 96 | none 97 | """ 98 | num_sum = sum(distribution) 99 | for i, data in enumerate(distribution): 100 | distribution[i] = data / float(num_sum) 101 | 102 | 103 | def compute_kl_divergence(dist_a, dist_b): 104 | """ 105 | Returen kl_divergence between 106 | Args: 107 | dist_a: list 108 | dist_b: list 109 | Returns: 110 | kl_divergence: float, kl_divergence 111 | """ 112 | nonzero_inds = dist_a != 0 113 | return np.sum(dist_a[nonzero_inds] * np.log(dist_a[nonzero_inds] / dist_b[nonzero_inds])) 114 | 115 | 116 | def threshold_distribution(distribution, target_bin=128): 117 | """ 118 | Returen the best cut off num of bin 119 | Args: 120 | distribution: list, activations has been processed by histogram and normalize,size is 2048 121 | target_bin: int, the num of bin that is used by quantize, Int8 default value is 128 122 | Returns: 123 | target_threshold: int, num of bin with the minimum KL 124 | """ 125 | target_threshold = target_bin 126 | min_kl_divergence = 1000 127 | length = distribution.size 128 | 129 | quantize_distribution = np.zeros(target_bin) 130 | 131 | threshold_sum = 0.0 132 | threshold_sum = sum(distribution[target_bin:]) 133 | 134 | for threshold in range(target_bin, length): 135 | t_distribution = copy.deepcopy(distribution[:threshold]) 136 | t_distribution[threshold-1] = t_distribution[threshold-1] + threshold_sum 137 | threshold_sum = threshold_sum - distribution[threshold] 138 | 139 | # ************************ threshold ************************ 140 | quantize_distribution = np.zeros(target_bin) 141 | num_per_bin = threshold / target_bin 142 | for i in range(0, target_bin): 143 | start = i * num_per_bin 144 | end = start + num_per_bin 145 | 146 | left_upper = (int)(math.ceil(start)) 147 | if(left_upper > start): 148 | left_scale = left_upper - start 149 | quantize_distribution[i] += left_scale * distribution[left_upper - 1] 150 | 151 | right_lower = (int)(math.floor(end)) 152 | if (right_lower < end): 153 | right_scale = end - right_lower 154 | quantize_distribution[i] += right_scale * distribution[right_lower] 155 | 156 | for j in range(left_upper, right_lower): 157 | quantize_distribution[i] += distribution[j] 158 | # ************************ threshold ************************ 159 | 160 | # ************************ quantzie ************************ 161 | expand_distribution = np.zeros(threshold, dtype=np.float32) 162 | 163 | for i in range(0, target_bin): 164 | start = i * num_per_bin 165 | end = start + num_per_bin 166 | 167 | count = 0 168 | 169 | left_upper = (int)(math.ceil(start)) 170 | left_scale = 0.0 171 | if (left_upper > start): 172 | left_scale = left_upper - start 173 | if (distribution[left_upper - 1] != 0): 174 | count += left_scale 175 | 176 | right_lower = (int)(math.floor(end)) 177 | right_scale = 0.0 178 | if (right_lower < end): 179 | right_scale = end - right_lower 180 | if (distribution[right_lower] != 0): 181 | count += right_scale 182 | 183 | for j in range(left_upper, right_lower): 184 | if (distribution[j] != 0): 185 | count = count + 1 186 | 187 | expand_value = quantize_distribution[i] / count 188 | 189 | if (left_upper > start): 190 | if (distribution[left_upper - 1] != 0): 191 | expand_distribution[left_upper - 1] += expand_value * left_scale 192 | if (right_lower < end): 193 | if (distribution[right_lower] != 0): 194 | expand_distribution[right_lower] += expand_value * right_scale 195 | for j in range(left_upper, right_lower): 196 | if (distribution[j] != 0): 197 | expand_distribution[j] += expand_value 198 | # ************************ quantzie ************************ 199 | 200 | kl_divergence = compute_kl_divergence(t_distribution, expand_distribution) 201 | 202 | if kl_divergence < min_kl_divergence: 203 | min_kl_divergence = kl_divergence 204 | target_threshold = threshold 205 | 206 | return target_threshold 207 | 208 | 209 | def net_forward(net, image_path, transformer): 210 | """ 211 | network inference and statistics the cost time 212 | Args: 213 | net: the instance of Caffe inference 214 | image_path: a image need to be inference 215 | transformer: 216 | Returns: 217 | none 218 | """ 219 | # load image 220 | image = caffe.io.load_image(image_path) 221 | # transformer.preprocess the image 222 | net.blobs['data'].data[...] = transformer.preprocess('data',image) 223 | # net forward 224 | start = time.clock() 225 | output = net.forward() 226 | end = time.clock() 227 | print("%s forward time : %.3f s" % (image_path, end - start)) 228 | 229 | 230 | def file_name(file_dir): 231 | """ 232 | Find the all file path with the directory 233 | Args: 234 | file_dir: The source file directory 235 | Returns: 236 | files_path: all the file path into a list 237 | """ 238 | files_path = [] 239 | 240 | for root, dir, files in os.walk(file_dir): 241 | for name in files: 242 | file_path = root + "/" + name 243 | print(file_path) 244 | files_path.append(file_path) 245 | 246 | return files_path 247 | 248 | 249 | def network_prepare(net, mean, norm): 250 | """ 251 | instance the prepare process param of caffe network inference 252 | Args: 253 | net: the instance of Caffe inference 254 | mean: the value of mean 255 | norm: the value of normalize 256 | Returns: 257 | none 258 | """ 259 | print("Network initial") 260 | 261 | img_mean = np.array(mean) 262 | 263 | # initial transformer 264 | transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape}) 265 | # convert shape from RBG to BGR 266 | transformer.set_transpose('data', (2,0,1)) 267 | # load meanfile 268 | transformer.set_mean('data', img_mean) 269 | # resize image data from [0,1] to [0,255] 270 | transformer.set_raw_scale('data', 255) 271 | # convert RGB -> BGR 272 | transformer.set_channel_swap('data', (2,1,0)) 273 | # normalize 274 | transformer.set_input_scale('data', norm) 275 | 276 | return transformer 277 | 278 | 279 | def weight_quantize(net, net_file): 280 | """ 281 | CaffeModel convolution weight blob Int8 quantize 282 | Args: 283 | net: the instance of Caffe inference 284 | net_file: deploy caffe prototxt 285 | Returns: 286 | none 287 | """ 288 | print("\nQuantize the kernel weight:") 289 | 290 | # parse the net param from deploy prototxt 291 | params = caffe_pb2.NetParameter() 292 | with open(net_file) as f: 293 | text_format.Merge(f.read(), params) 294 | 295 | for i, layer in enumerate(params.layer): 296 | if i == 0: 297 | if layer.type != "Input": 298 | raise ValueError("First layer should be input") 299 | 300 | # find the convolution 3x3 and 1x1 layers to get out the weight_scale 301 | if(layer.type == "Convolution" or layer.type == "ConvolutionDepthwise"): 302 | kernel_size = layer.convolution_param.kernel_size[0] 303 | if(kernel_size == 3 or kernel_size == 1): 304 | layer_name = layer.name 305 | # create the weight param scale name 306 | weight_name = layer.name + "_param_0" 307 | weight_data = net.params[layer_name][0].data 308 | # find the blob threshold 309 | max_val = np.max(weight_data) 310 | min_val = np.min(weight_data) 311 | threshold = max(abs(max_val), abs(min_val)) 312 | weight_scale = QUANTIZE_NUM / threshold 313 | print("%-30s max_val : %-10f scale_val : %-10f" % (weight_name, max_val, weight_scale)) 314 | save_str = weight_name + " " + str(weight_scale) 315 | save_temp.append(save_str) 316 | 317 | # find the top blob name on every layer 318 | top_blob = layer.top[0] 319 | if top_blob not in top_blob_names: 320 | top_blob_names.append(layer.top[0]) 321 | 322 | return None 323 | 324 | 325 | def distribution_num(distribution): 326 | return sum(distribution) 327 | 328 | 329 | def activation_quantize(net, transformer, images_files): 330 | """ 331 | Activation Int8 quantize, optimaize threshold selection with KL divergence, 332 | given a dataset, find the optimal threshold for quantizing it. 333 | Ref: http://on-demand.gputechconf.com/gtc/2017/presentation/s7310-8-bit-inference-with-tensorrt.pdf 334 | Args: 335 | net: the instance of Caffe inference 336 | transformer: 337 | images_files: calibration dataset 338 | Returns: 339 | none 340 | """ 341 | print("\nQuantize the Activation:") 342 | 343 | blob_num = len(top_blob_names) 344 | max_vals = [0 for x in range(0, blob_num)] 345 | distribution_intervals = [0 for x in range(0, blob_num)] 346 | distributions = [] 347 | 348 | # run float32 inference on calibration dataset to find the activations range 349 | for i , image in enumerate(images_files): 350 | net_forward(net, image, transformer) 351 | print("loop stage 1 : %d" % (i)) 352 | # find max threshold 353 | for i, blob_name in enumerate(top_blob_names): 354 | blob = net.blobs[blob_name].data[0].flatten() 355 | max_val = np.max(blob) 356 | min_val = np.min(blob) 357 | max_vals[i] = max(max_vals[i], max(abs(max_val), abs(min_val))) 358 | 359 | # calculate statistic blob scope and interval distribution 360 | for i, blob_name in enumerate(top_blob_names): 361 | distribution_intervals[i] = STATISTIC * max_vals[i] / INTERVAL_NUM 362 | distribution = [0 for x in range(0, INTERVAL_NUM)] 363 | distributions.append(distribution) 364 | print("%-20s max_val : %-10.8f distribution_intervals : %-10.8f" % (blob_name, max_vals[i], distribution_intervals[i])) 365 | 366 | # for each layers 367 | # collect histograms of activations 368 | print("\nCollect histograms of activations:") 369 | for i, image in enumerate(images_files): 370 | net_forward(net, image, transformer) 371 | print("loop stage 2 : %d" % (i)) 372 | start = time.clock() 373 | for i, blob_name in enumerate(top_blob_names): 374 | blob = net.blobs[blob_name].data[0].flatten() 375 | add_to_distribution(blob, distributions[i], distribution_intervals[i]) 376 | end = time.clock() 377 | print("add cost %.3f s" % (end - start)) 378 | 379 | # calculate threshold 380 | for i, distribution in enumerate(distributions): 381 | # normalize distributions 382 | normalize_distribution(distribution) 383 | 384 | distribution = np.array(distribution) 385 | 386 | # pick threshold which minimizes KL divergence 387 | threshold_bin = threshold_distribution(distribution) 388 | threshold = (threshold_bin + 0.5) * distribution_intervals[i] 389 | 390 | # get the activation calibration value 391 | calibration_val = QUANTIZE_NUM / threshold 392 | 393 | # save the calibration value with it's layer name 394 | print("%-20s bin : %-8d threshold : %-10f interval : %-10f scale : %-10f" % (top_blob_names[i], threshold_bin, threshold, distribution_intervals[i], calibration_val)) 395 | blob_str = top_blob_names[i] + " " + str(calibration_val) 396 | save_temp.append(blob_str) 397 | 398 | return None 399 | 400 | 401 | def usage_info(): 402 | """ 403 | usage info 404 | """ 405 | print("Input params is illegal...╮(╯3╰)╭") 406 | print("try it again:\n python caffe-int8-scale-tools.py -h") 407 | 408 | 409 | def main(): 410 | """ 411 | main function 412 | """ 413 | print(args) 414 | 415 | if args.proto == None or args.model == None or args.mean == None or args.images == None: 416 | usage_info() 417 | return None 418 | 419 | # deploy caffe prototxt path 420 | net_file = args.proto 421 | 422 | # trained caffemodel path 423 | caffe_model = args.model 424 | 425 | # mean value 426 | mean = args.mean 427 | 428 | # norm value 429 | norm = 1.0 430 | if args.norm != 1.0: 431 | norm = args.norm[0] 432 | 433 | # calibration dataset 434 | images_path = args.images 435 | 436 | # the output calibration file 437 | calibration_path = args.output 438 | 439 | # default use CPU to forwark 440 | if args.gpu != 0: 441 | caffe.set_device(0) 442 | caffe.set_mode_gpu() 443 | 444 | # initial caffe net and the forword model(GPU or CPU) 445 | net = caffe.Net(net_file,caffe_model,caffe.TEST) 446 | 447 | # prepare the cnn network 448 | transformer = network_prepare(net, mean, norm) 449 | 450 | # get the calibration datasets images files path 451 | images_files = file_name(images_path) 452 | 453 | # quanitze kernel weight of the caffemodel to find it's calibration table 454 | weight_quantize(net, net_file) 455 | 456 | # quantize activation value of the caffemodel to find it's calibration table 457 | activation_quantize(net, transformer, images_files) 458 | 459 | # save the calibration tables,best wish for your INT8 inference have low accuracy loss :) 460 | calibration_file = open(calibration_path, 'w') 461 | for data in save_temp: 462 | calibration_file.write(data + "\n") 463 | 464 | calibration_file.close() 465 | print("\nCaffe Int8 Calibration table create success,best wish for your INT8 inference has a low accuracy loss...\(^▽^)/") 466 | 467 | if __name__ == "__main__": 468 | main() 469 | -------------------------------------------------------------------------------- /classification-dev/googlenet_v1.table: -------------------------------------------------------------------------------- 1 | conv2/3x3_reduce_param_0 115.001241494 2 | conv2/3x3_param_0 164.772547138 3 | inception_3a/1x1_param_0 116.860034164 4 | inception_3a/3x3_reduce_param_0 79.1881406724 5 | inception_3a/3x3_param_0 231.357912988 6 | inception_3a/5x5_reduce_param_0 166.105403449 7 | inception_3a/pool_proj_param_0 141.278502738 8 | inception_3b/1x1_param_0 182.756980439 9 | inception_3b/3x3_reduce_param_0 188.983437856 10 | inception_3b/3x3_param_0 210.638443224 11 | inception_3b/5x5_reduce_param_0 191.872441547 12 | inception_3b/pool_proj_param_0 228.95841076 13 | inception_4a/1x1_param_0 230.266832058 14 | inception_4a/3x3_reduce_param_0 213.669944661 15 | inception_4a/3x3_param_0 195.305015429 16 | inception_4a/5x5_reduce_param_0 249.757731161 17 | inception_4a/pool_proj_param_0 188.650596899 18 | inception_4b/1x1_param_0 193.974176486 19 | inception_4b/3x3_reduce_param_0 162.319440983 20 | inception_4b/3x3_param_0 221.563688996 21 | inception_4b/5x5_reduce_param_0 263.254995677 22 | inception_4b/pool_proj_param_0 346.335338297 23 | inception_4c/1x1_param_0 165.722279791 24 | inception_4c/3x3_reduce_param_0 167.280460819 25 | inception_4c/3x3_param_0 314.666309474 26 | inception_4c/5x5_reduce_param_0 307.302805062 27 | inception_4c/pool_proj_param_0 292.606042179 28 | inception_4d/1x1_param_0 168.708545564 29 | inception_4d/3x3_reduce_param_0 241.418758023 30 | inception_4d/3x3_param_0 403.851433463 31 | inception_4d/5x5_reduce_param_0 283.655791452 32 | inception_4d/pool_proj_param_0 315.300139389 33 | inception_4e/1x1_param_0 170.752777459 34 | inception_4e/3x3_reduce_param_0 331.358737323 35 | inception_4e/3x3_param_0 321.119723175 36 | inception_4e/5x5_reduce_param_0 288.571297664 37 | inception_4e/pool_proj_param_0 239.612791756 38 | inception_5a/1x1_param_0 199.742036849 39 | inception_5a/3x3_reduce_param_0 332.047338615 40 | inception_5a/3x3_param_0 815.585545841 41 | inception_5a/5x5_reduce_param_0 267.603533506 42 | inception_5a/pool_proj_param_0 355.915136871 43 | inception_5b/1x1_param_0 356.912443144 44 | inception_5b/3x3_reduce_param_0 372.704808957 45 | inception_5b/3x3_param_0 1142.01213682 46 | inception_5b/5x5_reduce_param_0 500.777341721 47 | inception_5b/pool_proj_param_0 379.110336575 48 | conv2/3x3_reduce 0.915692808466 49 | conv2/3x3 0.374387932296 50 | inception_3a/1x1 0.915694621405 51 | inception_3a/3x3_reduce 0.915694621405 52 | inception_3a/3x3 0.351444458242 53 | inception_3a/5x5_reduce 0.915694621405 54 | inception_3a/pool_proj 0.915694621405 55 | inception_3b/1x1 0.187413273886 56 | inception_3b/3x3_reduce 0.187413273886 57 | inception_3b/3x3 0.26855407598 58 | inception_3b/5x5_reduce 0.187413273886 59 | inception_3b/pool_proj 0.178968517419 60 | inception_4a/1x1 0.23007774904 61 | inception_4a/3x3_reduce 0.23007774904 62 | inception_4a/3x3 0.180299177675 63 | inception_4a/5x5_reduce 0.23007774904 64 | inception_4a/pool_proj 0.210159959915 65 | inception_4b/1x1 0.222936949475 66 | inception_4b/3x3_reduce 0.222936949475 67 | inception_4b/3x3 0.317825212541 68 | inception_4b/5x5_reduce 0.222936949475 69 | inception_4b/pool_proj 0.17117225639 70 | inception_4c/1x1 0.336304619769 71 | inception_4c/3x3_reduce 0.336304619769 72 | inception_4c/3x3 0.329287581247 73 | inception_4c/5x5_reduce 0.336304619769 74 | inception_4c/pool_proj 0.248558637563 75 | inception_4d/1x1 0.316643235071 76 | inception_4d/3x3_reduce 0.316643235071 77 | inception_4d/3x3 0.363013449959 78 | inception_4d/5x5_reduce 0.316643235071 79 | inception_4d/pool_proj 0.264712104589 80 | inception_4e/1x1 0.403461089168 81 | inception_4e/3x3_reduce 0.403461089168 82 | inception_4e/3x3 0.819119193444 83 | inception_4e/5x5_reduce 0.403461089168 84 | inception_4e/pool_proj 0.3234631927 85 | inception_5a/1x1 0.636233639466 86 | inception_5a/3x3_reduce 0.636233639466 87 | inception_5a/3x3 0.846125123846 88 | inception_5a/5x5_reduce 0.636233639466 89 | inception_5a/pool_proj 0.515883878636 90 | inception_5b/1x1 0.791271305431 91 | inception_5b/3x3_reduce 0.791271305431 92 | inception_5b/3x3 1.35745862003 93 | inception_5b/5x5_reduce 0.791271305431 94 | inception_5b/pool_proj 0.70616444446 95 | -------------------------------------------------------------------------------- /classification-dev/mobilenet_v1.table: -------------------------------------------------------------------------------- 1 | conv1_param_0 156.639840536 2 | conv2_1/dw_param_0 0 72.1291432409 149.919382216 247.459610328 174.235629611 150.13176085 63.6101305164 119.918502265 0 0 0 135.378873665 91.7824651437 0 0 168.873320426 169.168022632 201.568614006 189.509074697 204.264122386 365.864018575 193.888303858 0 171.289626192 138.189660082 0 0 0 142.681533662 0 316.949893002 139.107203821 3 | conv2_1/sep_param_0 100.345310403 4 | conv2_2/dw_param_0 532.225148018 362.789836739 0 354.656744009 476.643207117 357.446052894 342.330413966 342.355908638 329.233026576 326.08016418 445.073837291 575.536253211 523.241806076 363.721168885 407.948692092 440.047082231 444.488991906 765.897664526 396.299719381 523.063806743 0 0 0 475.732604565 0 386.344972544 496.619603747 0 0 558.177980748 350.906105109 355.089923289 501.953064809 0 328.815968553 424.278407198 406.471325993 437.9321272 313.476557668 344.940862175 316.267070088 521.271130506 255.897909209 549.433257837 508.117661781 658.414828519 0 326.38095621 431.332458808 0 403.314530911 394.589660401 323.462216655 0 541.841966458 411.173303515 369.845542955 357.724416872 364.36237111 365.442055347 495.067332857 514.383243088 271.295931948 342.438551776 5 | conv2_2/sep_param_0 172.178812333 6 | conv3_1/dw_param_0 307.751722369 273.557398253 381.571041447 177.479181181 273.682259542 415.604750094 214.141306906 343.107683104 462.515794833 536.694267745 227.801738781 462.44828634 330.726879359 364.561740946 98.04308453 292.31883469 287.770341944 340.529211835 396.819462961 351.082542211 312.328248126 262.088909092 413.583766571 257.193065139 315.398408107 282.951632242 370.131892116 222.054779739 197.070234928 524.880751035 0 320.8735517 272.120333861 298.538979557 207.333981467 271.544058168 590.07178405 262.665663444 226.495194579 156.591893375 413.321740625 148.722307874 280.512107424 173.584592233 356.114146047 512.247382019 174.920499706 536.289827193 280.49602532 377.035870678 363.586920296 337.371438705 350.797030113 445.910365737 216.994180068 595.440211599 933.840156561 920.795364792 344.308628358 197.938060726 553.193554264 420.999410105 0 285.654531201 359.777213357 296.126583969 281.854369421 182.640631833 345.176539277 387.91897075 143.433156998 373.82867375 186.532676606 514.007192147 311.579353496 308.461051101 0 490.452452819 215.181596177 401.008292191 515.062500604 217.056716147 605.304654436 449.000815941 135.823949948 305.627108083 0 315.059312357 311.237640549 800.548186932 570.477899838 382.330116575 141.529757028 835.214955191 286.153119861 507.81566683 884.16466651 294.692206765 469.511883196 447.307473468 342.2395771 247.262439502 399.006117283 280.720603015 318.27877414 305.934289154 155.233803463 0 535.029937179 377.773039707 249.361989141 601.122946298 327.904985866 225.832075011 165.260204886 422.508767273 0 369.751293158 273.149701532 402.70410863 282.519446374 187.462167588 241.25283429 286.528941965 224.404908773 194.85102058 376.469088928 652.362647261 7 | conv3_1/sep_param_0 128.391132187 8 | conv3_2/dw_param_0 675.697910637 433.92794989 631.813706998 672.327946869 540.568948433 606.640059847 639.343762284 532.342230536 590.692014939 655.469523812 666.332182493 586.286982856 586.665570452 668.721995997 791.972592709 552.108193912 586.871312733 738.075564618 645.300256861 543.352846196 904.058260602 774.673813481 693.467039789 509.238827318 598.899425353 0 752.508654336 600.894212607 558.503082052 705.728635457 596.681181188 512.591788043 534.022277386 460.152640552 777.4509021 982.202867819 601.582167901 540.733469098 0 618.38419342 711.094004828 647.090253436 430.423700928 632.69571732 534.087968838 0 626.496355948 643.581218582 580.119585355 660.281393311 638.005111032 0 660.419894613 657.983779129 992.070228322 0 522.554530781 537.535322589 568.810243145 492.995977978 506.137539201 522.027987471 783.462591953 540.348578432 582.356255439 630.559222544 590.856554177 0 584.395500193 622.738078739 615.939947554 577.656127282 633.280168415 796.270739082 645.963605804 681.989190054 681.257453519 558.887927859 618.351262332 551.24569833 635.80553785 634.628342264 825.24546815 564.389850046 621.918263981 470.50163953 0 597.976165921 741.490281346 593.064966331 622.522339134 667.932946092 638.033959402 518.977136102 677.515811053 0 452.470019094 0 533.423834776 641.613424487 709.467596904 465.962028594 515.745395806 585.77490663 541.573475511 612.804689479 600.918234811 873.234167504 549.488518305 622.296479992 598.542801707 634.402067584 0 486.380093242 540.524791493 547.880356374 607.207408495 728.123779644 0 599.249985516 563.070223685 594.934744174 515.071496444 666.197803902 729.882640727 723.579796234 635.849984083 588.723082223 9 | conv3_2/sep_param_0 120.463218716 10 | conv4_1/dw_param_0 392.371915228 625.240951794 154.132992202 317.850531936 234.37363645 639.073857693 261.257386673 349.84068349 394.052306693 421.807217243 441.156214627 0 295.717136109 1103.35682471 410.681074159 402.860273213 546.21576047 442.751093679 393.413965443 170.541621651 452.353689966 529.446422287 418.113014781 269.824931168 549.823790742 488.8643413 570.348901529 0 447.781491194 722.662318338 437.850818407 769.373221798 0 298.040812039 1201.11544487 1111.97020386 168.398842369 0 360.769187997 656.4585565 364.919507036 703.980958507 496.559131305 676.764670621 354.527273493 322.42329146 329.181449931 529.22690855 416.095086628 0 397.317195296 551.343157616 1051.01779312 227.19920769 386.307007534 459.545710016 192.103990994 0 623.48241898 0 1189.18232562 0 283.691575849 207.605470705 404.65895417 232.677389574 399.138938502 395.484807003 272.088833473 438.374792548 354.316776068 0 226.222158896 387.626521419 371.688770077 428.327297596 387.694231048 1256.62227946 473.157215542 359.389320332 0 269.397242337 422.382253299 430.353630773 0 402.872498906 409.66678254 540.446300461 311.045540509 425.852898817 652.127743442 1084.95178505 472.39048819 869.292452872 301.668129248 312.946364457 219.45480287 244.036257491 269.670316725 211.766542478 807.435734977 332.254425627 290.493075571 416.918466241 433.997907113 402.493010797 381.927454616 467.528447408 351.119656182 546.795377262 472.007166814 0 940.513249336 331.020906283 997.882162026 289.057402893 340.954707423 251.280656189 1479.38165824 329.18836657 646.500348251 370.94497141 325.733334556 549.28864139 238.027465965 1168.61006779 0 339.6377302 314.078744243 213.943575778 612.483912572 305.98123248 284.268219938 474.518335835 1283.39157924 322.836411207 394.6561697 428.789490292 577.099216983 343.612587049 351.048270104 1303.96215357 0 316.759508424 1215.62306955 217.87077205 322.564892883 543.332755628 253.150883388 365.78751667 894.654858339 365.82472727 241.976365498 361.26024664 200.829540469 233.064424448 431.084707171 307.746721761 355.458119438 399.628943199 940.604695262 341.19501998 466.83726475 462.531809033 249.295439418 578.369693487 656.637546999 233.29810189 0 424.37592499 415.092034551 986.019789288 290.792717498 749.501880872 620.54776113 573.922788803 406.367097696 383.877038648 321.066036682 0 0 456.920992279 405.850506245 472.412482897 291.640006696 356.978369425 285.096571472 552.091885308 648.782526132 0 408.909590894 440.442999113 586.401706111 497.893343375 1603.80859697 239.778085472 0 316.942773907 499.157145004 419.665139861 210.852809014 305.863385262 494.578944607 320.073696495 453.65080746 411.732302734 437.700384545 535.406041072 371.305514228 457.116898814 554.698953847 660.467388301 428.234281451 1074.12999936 471.073345569 235.150031933 388.803158298 181.501907523 305.826288551 688.39850163 314.980880052 467.468236576 323.381974852 290.3502121 312.952937422 352.876239388 303.484827977 346.839517578 426.768679035 348.540289636 370.55712479 224.626627436 502.107014902 441.896212048 699.233504299 822.696037474 731.223919361 158.63233919 341.108388716 340.463970761 1013.70627513 292.046500318 0 222.440945798 418.341559037 488.295491788 366.642819612 605.042015394 461.741760859 423.606846637 373.912447858 488.070334876 388.751692835 281.097986238 246.026565985 394.218971539 11 | conv4_1/sep_param_0 206.227559089 12 | conv4_2/dw_param_0 845.875404958 749.590278751 942.712453439 682.982265257 765.405937293 698.379496017 1047.63090236 0 55687.6764578 720.083093695 0 703.033369155 881.397147805 959.723633579 868.062030783 766.695233807 769.52187977 814.584342553 886.019999324 684.749390337 695.680962007 602.001990892 767.092984677 834.026973393 848.895226179 868.214482191 721.759205095 735.487775923 507.810825726 825.998075055 900.044419996 786.220620988 795.629756022 653.604864275 798.704590307 842.350752023 1039.8945184 912.006406754 890.727903346 723.348151792 537.932782842 939.996981095 652.478614142 905.267326245 608.71713387 858.333999161 822.500646877 930.54371898 901.342653426 691.277368326 992.056371092 803.747022831 861.733984974 634.897909413 797.599173467 0 1043.59210302 811.517454219 644.344828832 815.688892977 874.235321101 906.491971112 0 838.851305128 657.443072543 919.336296167 861.976271755 1208.02111352 867.263685707 704.33654141 609.788215214 751.35083512 873.922645344 766.563385331 903.887690374 699.733067107 564.676208783 666.557049276 896.361979887 751.415819429 653.461540667 719.745474285 738.514489538 0 772.882498782 951.481460915 607.621346417 1063.92841636 0 0 734.174142983 943.417453547 910.295913422 834.980605678 653.692392746 576.04786371 816.527722033 701.05319937 861.832713464 773.151589978 895.321670347 887.398622344 913.157779598 773.857248157 754.664527625 898.374957731 920.214858097 594.215123102 871.583082904 690.563229961 929.558945964 875.511647841 868.395300238 819.741381166 607.516098508 820.481292249 890.598991324 832.428162217 777.580705459 851.003004682 655.247540561 750.576327943 739.761517119 815.896993651 943.398030014 907.299103658 729.487940692 752.003307718 837.787920214 872.774354594 642.009453694 723.405250775 974.422651977 767.28752489 797.388588138 961.181560004 729.431875207 0 0 1095.86476215 764.656254102 727.757202322 792.775127616 1042.42329146 868.88195432 735.395438919 625.073165272 1041.01118637 721.604720507 639.349133924 959.781995414 759.09946927 719.983757416 660.156040129 733.435383133 836.481766818 844.823876244 0 793.839845241 744.153952474 688.701503086 732.201837544 690.746244725 863.003929518 771.980609422 877.879701344 903.732516851 879.39865443 846.821505593 706.427187236 0 768.376614836 767.227847114 0 834.942569021 816.437457725 0 731.639468031 749.801436942 750.061888456 731.381106949 725.486410833 616.866555494 852.171875673 817.072703893 478.306385215 882.093815385 925.797464656 820.540220531 755.094777097 506363.977523 565.14758644 695.889313206 947.15880754 811.961614571 0 755.82072591 846.344534909 881.804962026 1050.28769241 682.463807018 830.080250328 625.581295155 733.093075049 765.907300406 0 725.874627655 716.384744887 660.292493786 595.095755469 788.804405919 731.424478771 871.727322455 848.626689024 597.552720084 619.707924076 729.692484194 0 836.161463402 593.179756838 875.844632693 931.439669059 746.38828678 651.205888495 631.907817719 1171.27988933 800.026520534 744.194498681 1012.57622432 0 625.604530576 980.283931365 783.888028737 744.953996249 629.207638263 772.178966169 0 764.300228434 610.014908059 961.547764562 772.589011861 676.371688723 618.892602401 1001.59234793 648.698825575 780.88343315 706.770710617 940.324704743 593.441369398 702.2410482 679.946685601 745.579812614 1094.84857342 850.561546567 773.262351787 272096.760958 13 | conv4_2/sep_param_0 197.963457842 14 | conv5_1/dw_param_0 431.074764652 536.091578366 483.946300679 358.505616745 523.606049503 430.484617904 563.602349548 511.694326303 760.22983869 523.432238521 261.384431891 477.288901493 858.50423887 513.151791694 1334.30174779 706.211133191 348.518369089 652.484158811 484.482857437 277.555808747 521.487889767 0 428.842220473 311.180094699 296.761825572 458.126338809 282.797019856 443.085542161 407.286721474 639.447134242 237.637673404 413.429888366 406.888657843 467.408553921 455.503692098 463.296911062 413.731854528 711.164257854 390.124236626 434.431233528 421.736376469 1498.62789625 330.181854697 0 0 686.654890089 1740.77518429 351.386685557 244.192067535 821.310699913 1389.798337 650.500305336 0 398.885369336 311.274811206 1722.1166137 424.640437239 281.227470589 340.846741246 416.767069731 566.272117023 467.473108256 295.962524954 635.212394162 240.095207542 484.147315129 258.423205017 607.241326542 607.216233745 348.326019877 314.524870558 434.586696592 488.459820708 997.903893907 1643.1273013 616.150034553 226.745988539 590.021824816 304.583002156 237.225250301 391.116742957 886.52642918 348.421056918 396.486587519 1367.94373901 383.102997614 324.911328159 324.191269381 0 491.608778237 459.5748513 290.913692747 359.135658632 335.84917585 482.38539776 390.92840988 405.454009215 346.563088743 404.808640198 437.051669456 558.680020373 333.282278069 0 581.83914273 785.440921986 453.283879794 1552.64998905 285.81663631 1634.52446467 230.652623821 333.325657183 0 547.151887347 0 467.030610282 305.906617574 452.236076835 0 648.616084486 361.58688971 0 518.131428247 448.895389698 486.104015268 348.825564988 376.156351483 406.557493261 299.544791139 512.16540775 338.834428985 268.425970207 880.762012473 341.401287585 431.188956804 519.129944543 336.129105243 462.767079457 329.124424165 488.393650466 450.004484191 546.568149938 0 377.851957387 679.159514986 722.677392437 322.463108868 424.169660982 0 218.650314993 470.074962467 216.616591823 1200.99239652 754.727947732 664.336729499 370.201924586 443.609612644 363.476176236 369.481004829 268.058856524 418.066007022 577.093355545 1133.73787545 269.783592226 513.851466819 317.41086237 585.50701582 527.183226979 904.997404623 662.121687605 576.689281676 268.050442696 349.746965966 698.882996656 364.621694284 523.801286268 439.310950061 0 247.484501572 332.58881013 638.518944989 1812.75893769 927.282982241 630.379757614 532.570326934 335.682504893 349.233531989 567.02465826 0 378.64032725 250.350278073 1055.90314681 383.300378091 231.63661937 378.864592457 368.226605834 422.369861433 751.702520455 290.006949284 387.729223626 1327.03900419 356.397085404 462.099163097 490.774863613 694.194041424 779.592367028 458.219787413 531.578532869 402.955851438 494.414890273 834.942732612 473.343792022 513.879784763 415.068099638 451.191721492 460.62222228 571.693357616 264.255114866 423.572488685 476.13120641 486.621752134 0 317.794354249 0 470.610443757 369.15264538 304.230106753 301.24409775 415.795950514 533.025339829 374.090058921 370.454203092 565.196045603 0 0 616.582280873 586.447503068 457.930698082 299.613258905 402.351642037 446.349258075 384.635292036 311.662665201 330.437627624 726.442655117 372.157659391 197.52550451 392.539258551 312.190275054 362.010109476 446.368240012 0 274.700035738 233.140445166 562.855216123 352.771017664 0 0 289.377313971 493.188713843 308.527111048 931.999571772 305.259216614 503.097493941 265.489087375 421.310123217 1957.13093572 331.625292411 272.877978242 252.536849504 378.921323601 1142.92250953 414.608696735 385.205707037 457.151666854 0 489.0091878 339.773429039 697.811067577 460.104249673 460.925688801 1398.87018415 259.14569893 501.315559864 571.869697734 431.451417871 407.318410222 390.773939325 327.489270136 226.396716147 309.436998602 556.299011056 720.377065727 640.115805123 306.451723752 510.484628786 473.150596094 453.39094355 604.672935916 296.499843206 1527.6090517 556.618509528 615.241964244 504.220662785 418.693277298 417.964316112 276.95857482 1271.95618459 367.14656146 379.860929994 387.374511045 605.893583916 489.67882824 364.339879331 328.574935161 0 718.130363169 567.851305332 292.871893503 343.22645782 494.230308727 319.18364494 326.857577997 0 401.181498635 544.459076736 466.504312475 554.449745548 957.204215603 333.742400108 79703.6370397 409.518559503 0 750.402391194 435.410122933 325.649349139 475.305775036 327.010178309 416.806814429 504.109479936 0 779.778959115 610.753022207 445.303916236 421.862796108 329.792428433 333.334417797 0 669.897541073 549.597406863 404.057943938 432.163353076 365.796151366 384.801833072 278.68958244 330.55371373 621.882868086 410.105110313 409.388691745 475.448956613 493.744427839 0 340.503063432 517.507713014 524.309512946 353.138017512 384.752150896 476.980063359 534.984497686 360.065549408 220.150257186 550.753152598 645.091649696 1293.03934148 418.0217982 0 1148.92548103 306.418450111 1644.81176585 662.555602052 394.624227821 559.2575239 1925.93674469 782.314047272 385.457726638 487.437084729 349.358935094 394.538587709 371.219411579 422.156425255 324.826503275 347.414205035 0 231.414198777 706.106402461 410.558458014 256.794225612 405.820050292 406.276556672 562.248451872 404.923843874 473.557668646 418.397049885 381.206221799 254.682474057 0 379.785638774 460.931272622 393.471904387 441.033944215 662.961463284 453.961016186 356.436255783 0 362.910701618 421.662221617 535.202933095 573.300001117 556.88968419 481.061423255 277.440140586 692.111555755 0 302.370877687 342.322686557 357.727299697 297.839170828 637.575989308 491.104490134 303.543584201 555.689154575 1303.89691982 331.3925197 414.421810577 519.488420407 248.429690774 394.289514812 566.629395523 293.678385029 313.053013892 359.151397946 0 725.050987959 427.636388742 341.634451759 305.211991771 419.800369735 729.580174052 624.695780055 503.440734723 255.935009736 628.023024297 446.350193108 911.687784139 638.716477229 544.095297977 303.990988871 985.235924265 681.618573546 418.802114387 462.882039082 355.734730913 439.783865676 534.034189664 626.550057801 432.028276507 622.637536307 471.986673566 0 442.885226724 563.826693741 350.583669287 390.67928833 0 481.977786381 444.943021174 431.073848916 580.414464309 343.518271693 389.521156052 1432.95172537 0 363.207767955 395.406791022 582.004986238 674.408055207 258.547100248 311.481104774 332.003619042 432.3099166 276.705130906 424.780248007 338.057524819 585.255485277 823.693808108 416.612321086 497.998832779 587.680814184 716.168335183 420.306839044 286.628040134 315.565168594 536.658243212 324.309523106 363.692144669 489.484158721 444.390168677 417.590862927 314.183387318 368.910407261 334.998694174 368.776162874 315.368391263 15 | conv5_1/sep_param_0 199.589322533 16 | conv5_2/dw_param_0 544.677939594 1279.84374985 356.148045283 599.838289355 0 256.533760058 1468.14795851 1221.29293589 698.968054055 458.875836632 732.339308513 863.023591819 644.618961216 471.472624931 575.629156216 507.326093833 608.901701052 1591.79208824 528.151537663 1521.94009598 401.549964994 617.440007047 482.537193604 838.823482262 413.993355181 521.523533871 1339.19593459 917.069596385 432.925664488 1789.45867569 333.345890727 549.413458874 722.030140525 345.824034738 646.497552961 1174.10765243 1374.77625603 0 1062.17922532 460.292504667 506.294488234 339.631666768 627.36574301 484.841591169 479.861006259 515.875103981 596.323021275 455.382878237 773.492534778 790.367769918 333.559979561 493.760732433 463.312727536 1349.25385085 492.686136501 1145.2672443 715.113204911 307.909602857 501.792650304 456.590239301 271.393171439 442.536328027 0 440.493625917 498.15770395 1263.82509952 1649.94104249 518.250584724 427.680594395 498.295524814 535.901057676 558.9113111 554.356413257 651.826649265 628.830024499 403.732975791 507.364328493 846.90447531 849.380663762 886.348490691 843.331703103 461.222317058 443.918721888 807.072700622 553.241780625 281.199652933 359.277786213 979.737506827 584.827544259 1088.58310317 589.27120934 787.213184146 1252.41879196 282.00910958 306.772203358 541.223920592 567.315925615 920.741051058 547.051163713 492.463344042 633.156813558 367.586613553 519.489496991 1328.37980687 578.81873523 708.493822394 463.32854509 523.382126506 2048.31267076 1326.42519746 505.80087833 361.53618087 575.759426404 1243.14039584 658.063236823 434.986120173 983.26395022 457.662124498 777.775633743 1106.80909659 949.55138857 489.641243446 765.566887958 533.519802703 0 780.604429707 710.477217206 650.14382165 197.515964705 601.774754839 651.555020486 1627.57452312 480.313599854 424.037931051 0 311.101673888 394.429472365 502.545191842 293.617073477 767.361098814 516.989875029 578.390299892 762.970948412 0 721.511560633 1128.6595914 983.369004563 446.055057603 882.869223915 363.886648806 352.113738988 454.478085955 611.977229357 437.050638503 412.711889154 599.128285177 433.768763084 1018.20601382 988.82220873 575.788443852 962.875137294 470.627127363 1065.68491043 1012.44535297 1242.74278752 1222.07764732 640.88552203 595.058610457 381.61679553 0 567.688966959 1461.28726524 0 538.757632746 437.102326603 481.308207436 863.667611995 899.16655568 948.844800195 476.974511031 601.766554426 583.042277894 483.388284148 721.651034498 333.629627449 273.87723712 590.829192869 514.877673003 208246.045987 696.01939715 1116.77951677 757.460891538 477.427985101 588.003836764 578.389514859 412.294738667 1370.74258406 535.428408843 896.768099529 402.122167429 223.217508447 717.826371575 572.608153333 269.466796668 639.033942671 770.180416215 278.000185924 1173.65324262 531.340651558 0 478.926129795 0 1670.6707446 557.529605212 693.525782712 507.80525857 1532.7748104 481.653270822 629.970141628 472.367657756 670.414161123 694.8976954 571.581250062 869.263903943 1249.65984496 524.9203844 686.01613503 508.387414282 0 876.804695811 1424.14072797 529.700419978 873.941912181 0 1004.33130153 515.019113322 1089.91450755 816.185542887 342.767731605 422.921584378 272.240912356 803.570983696 510.932869794 1346.83159062 729.53408568 730.773250563 997.72492619 0 954.338373227 668.91309234 977.601841687 0 1123.29694434 631.5697306 685.431427549 450.866012027 797.897630908 414.893846146 512.360520657 628.11934202 632.524892252 924.770124117 526.536110965 955.225292077 499.184392751 592.494150921 813.965473711 475.922281079 451.51106079 0 495.917758013 0 1094.66689015 519.990370433 297.70981017 0 900.172753475 841.739449499 398.893210318 521.999883385 764.677864945 615.272432943 769.499299509 1296.91080395 582346.219367 659.361914313 651.30845355 504.816600164 1186.61371669 455.982026996 0 501.167693916 798.588440817 1141.86066376 564.678229059 444.15816252 1189.84340849 0 416.188512689 0 408.075381199 372.091043717 720586.281743 730.814043681 546.754792109 606.455566384 407.542708126 989.522288624 477.502131968 450.646448867 514.443321951 581.413960265 393.429293083 1821.75130203 420.269864235 535.807094437 981.888745111 581.952489364 511.830703144 449.397091551 499.353500526 467.045658962 727.453762297 937.800094519 375.664290041 247.226204158 894.732719301 641.199793229 0 324.227774993 426.731413271 1426.33378081 523.748078107 394.783184329 904.601080916 0 763.798844447 1528.10329207 632.185769543 717.419295666 576.413962827 1194.02099281 369.596047848 521.598316163 379.746515377 610.015170027 568.133992506 509.228178084 542.947270034 544.354752573 424.563311939 385.22210807 551.807033608 413.346476885 799.704031959 458.562676464 259.473731819 540.620519705 1248.35641473 364.8121035 860.190389066 541.351598189 516.565323 354.540074705 841.715424842 309.686744817 194.821766281 308.176076239 599.149976147 815.89839957 0 481.889218963 646.476270316 594.264634807 1588.55429619 844.620764459 1135.71153963 438.947175184 602.854995732 600.128796765 1692.58620525 450.445001406 445.769218313 602.953473226 560.644462113 832.834390676 725.355636468 447.801818607 883.452545511 998.371007547 543.047699015 0 841.759401814 382.14435694 1603.02821626 701.671575751 384.204379451 705.93492179 562.107984847 263.467056485 475.017764233 1760.73887841 521.571726575 592.547331094 677.413010538 441.232862234 1259.37421844 342.850187536 675.738465552 651.167428708 1655.63934725 444.760007748 600.428722341 460.142156654 676.563154648 346.752620654 661.814276968 451.96023814 591.902481256 345.467618933 0 527.513797494 541.68981875 553.935762416 432.465181503 764.886077658 614.924798275 603.958380655 568.172473037 683.803673167 600.925056281 538.232683069 871.13506815 917.138084133 504.692004148 0 1548.27991322 512.009688914 0 0 457.798560432 757.756536728 228.408208735 451.659602269 451.530962741 850.461649363 772.501478369 700.892521613 518.318442072 811.907547151 492.273051187 287.534328313 411.87194179 736.371956744 454.64017868 312.092532146 1297.16592869 1441.7269356 723.573960313 437.78195229 661.7095583 0 822.830030135 538.769348524 937.651524691 756.151182581 344.13392999 651.876554711 622.412548117 1346.04116004 303.678108733 473.123332224 583.483067931 435.685366469 629.606123416 690.719206163 294.399345628 407.525988286 303.293066663 339.676010612 276693.276279 743.230061048 0 639.35397807 633.955047555 422.287699994 1379.68058325 194618.231112 867.505560373 281.593955746 500.05020717 905.032577591 1325.30644206 442.936646759 516.111524758 442.501220265 461.039236469 958.118767112 570.07003102 795.497272778 776.807349296 591.485168777 463.825029192 490.930232912 691.176122949 445.639158235 397.690952861 500.070627869 484.603680584 631.710962051 535.329264882 397.695221018 17 | conv5_2/sep_param_0 233.45947638 18 | conv5_3/dw_param_0 919.443110951 1903.42531505 980.048225844 499.982326205 0 844.119933047 669.178793186 621.211382446 897.641568174 534.461542533 591.362990716 1020.0863121 0 841.369754671 644.834629637 527.328475559 840.098788096 355.960683228 1503.97211304 566.87968343 573.117960484 941.575782215 1241.2758252 278.396636725 1163.62363167 490.511051717 429.490799245 0 303.553011526 524.891580112 428.693556357 415.683439297 1122.66098989 274.212222271 0 1863.23120761 528.350343301 1727.93144414 521.605051771 519.852560139 391.91593413 0 405.48965759 580.642584665 541.64823235 919.56999195 923.847294881 1157.8504905 649.735418028 951.125005133 591.298659245 359.547392505 642.165457242 913.556741952 799.966597262 835.232389373 561.369575334 308.871468934 1382.51201357 1712.15598918 722.951286779 521.161033797 793.982723932 834.619928562 0 0 594.62957706 961.918048053 601.004128841 250.129564305 0 671.86276345 915.840925612 865.857398508 1360.88299177 518.617124646 565.061931868 1112.32096891 0 0 773.126972759 868.138691939 815.381894402 414.689712916 983.466250886 1259.64718043 341.203762022 1231.83137715 557.988793119 479.287513717 690.07661424 724.483779069 513.292965346 590.116031384 616.496737412 422.743108199 0 495.070898762 546.578525342 974.463874183 539.504782846 678809.334102 1568.20847755 1386.40704901 1396.7164166 749.157453386 444.491170962 693.363685435 0 929.8743578 852.069725782 278.72343196 283.836278368 356.072249677 400.584364473 0 1043.4776846 435.565174516 0 274.938908571 1145.48012337 593.817795445 1311.59814384 950.176184528 929.988202013 462.677443549 480.667051148 46619.8161511 736.376092231 830.521659538 1219.53768918 1562.92090602 0 1300.12459637 506.764895405 746.042991968 454.61631572 549.376379504 1835.54445864 404.210406181 544.241189026 515.209492767 1505.68494307 557.676295237 723.091011132 563.460236764 1554.67957596 1120.33918302 1502.50762032 790.500456288 268243.316164 1654.06569296 543.060986238 409.075829129 508.532562561 1523.85144978 798.327527051 1276.16698978 1245.18001211 1644.09547832 967.030070905 725.387245213 760.98513187 1207.24947068 229.165904235 606.132520076 956.0381418 475.420366436 0 866.31091877 1649.25746155 543.541458208 434.907366256 779.155124428 410.259919776 839.218541683 595.832466074 279.360739806 795.46720285 453.130461708 692.423348314 1726.26656041 388.878164074 521.987830569 566.121585183 942.712036344 505.635775563 599.976665548 303.949705441 592.57472817 1578.07764586 320.666070797 882.076561009 490.603720744 705.754641117 411.879784123 1706.86275355 1727.27326172 1659.2151454 561.923200413 734.23422904 821.911940311 0 1667.08898521 557.061814864 254.823849465 676.909098851 577.110471279 336.41226364 321.652284469 513.826063879 551.195894357 645.484359313 1541.15062601 1354.31051592 453.304613419 763.540942954 660.756438398 626.905154512 609.360559288 1445.03902652 681.821367309 608.65775174 1403.11974208 522.187109171 661.035282307 1190.50256607 653.582158828 1082.67011854 690.687300024 497.667968155 869.802216015 675.82955774 0 540.831502001 588.034223284 1071.66741342 703.379984316 432.609358705 505.941279455 502.812024928 1094.68587126 735.152490806 503.747104444 1587.51940374 1564.87343242 732.529147315 725.963599715 1183.7918906 576.36675718 1327.93005438 771.083543177 748.154897829 433.135867585 396.558719715 297.213906255 1655.36633386 900.091376298 1738.61292408 600.871802088 966.720751959 477.803021382 507.082445569 409.427946318 488.432387552 509.896564682 1600.40442406 540.478619553 774.585314187 963.279649202 905.855501242 1670.07737968 843.599404133 423.685983756 688.147935696 745.096623515 933.649061495 2079.56498993 1582.90541014 581.70181927 566.247248549 793.105704457 567.166650052 710.462470093 786.269435312 557.796302081 700.578988464 0 989.2704075 0 412.138390033 786.855970931 1412.4813715 462.171682369 431.184244854 837.265298676 512.046540894 246350.364084 470.799437879 615.413800989 474.11235363 1301.55294308 664.894545241 396.131143609 537.785231584 11036.3727209 524.839636936 455.312230376 479.036065527 678.121723389 1737.82698022 1760.36865596 1133.1814941 639.060631995 1666.10950549 1550.10620764 453.427269907 571.179841088 1926.48896886 524.068133817 356.978429233 989.962153239 820.402313191 768.608957801 262.417846971 247.542725244 499.775628699 0 499.526587514 742.507902692 948.628403982 552.394251768 667.452756801 1921.21260637 353.266709426 437.304758546 261.598948549 0 1671.92318608 850.996886704 297.889243193 561.94172526 1513.73144438 4072.66835799 447.354947481 373.385728794 1365.82684061 1574.58707637 489.709721859 603.131274776 566.030484981 720.713510648 410.033765958 778.885540142 590.257234141 382043.136566 586.23931572 533.689985215 1404.04144667 626.257664557 460.70045458 449.841405784 472.100820596 292.811119396 583.651129049 723.881856804 1592.71200747 248.888977769 728.574237317 785.333518606 1535.00551941 401.900686673 703.486346572 423.225054164 990.562410834 1166.26734041 0 835.516022129 1606.09770448 306.434865673 557.874180861 502.442270013 407.346249002 954.867630883 598.79090859 550.645335416 1944.9412074 924.885331407 579.373693137 433.550182107 378.421368 739.109786628 620.316469619 1287.96938672 389.765950342 494.210419535 414.393036615 2071.89139207 615.129620342 0 694.081200121 826.749168796 1568.96542591 567.315434697 651.303277247 503.234852929 562.162338891 824.049802516 1938.80871906 482.416907069 1232.93844809 0 742.866183993 883.854106646 562.349358517 729.646381632 850.140982651 534.323392696 273.272841466 592.62359601 791.296629405 583.103069937 1263.44046603 595.464465352 317.22292276 0 267.244027414 1803.68484243 517.808324496 314.698214672 504.551518824 1462.78480894 705.855643068 510.922486504 1233.16456175 1207.90751687 1588.36556264 629.324440255 548.043331997 723.451304909 1250.21950111 868.249507595 392.277138626 254.593157849 474.584499151 535.933812739 1839.13839643 565.780983781 473.485635975 242.976036594 459.801218522 440.721043679 1223.24036698 1609.69579244 1724.21324521 505.090820304 538.292002888 275.105166305 248.735770961 728.606750082 931.448219906 793.067402151 449.602678153 484.48864103 1684.60360366 1311.66041598 691.02330198 825.959891816 794.586161873 603.511978184 471.785912981 797.080820298 1164.16841298 393.266525168 1901.73662875 1574.15854592 591.310433155 526.363437824 348.214274601 272.885160097 539.892097162 1215.57174944 686.892190377 664.927847885 1818.45420508 674.541549143 242.930659565 644.36007665 1041.02167661 720.923955982 0 1358.40671047 1395.8869428 846.123554832 771.147310968 1480.13301984 921.104060084 368.101697829 1525.96634927 451.245900848 839.193090729 1729.33913931 1222.08474427 1587.06710835 614.105411984 407.133680537 658.005572201 472.763837791 846.766229521 1402.18771224 871.311403944 1200.65917551 19 | conv5_3/sep_param_0 223.146231084 20 | conv5_4/dw_param_0 375.10376754 729.92714779 561.305356226 1027.1047963 1310.40571947 346.670307722 265.652339849 1785.95504615 0 575.011700712 660.107466423 835.582700506 755.804773666 1487.31147606 860.378041872 320.9281406 742.259198938 664.811043936 4214.80319939 613.29818026 466.870253649 650.870053653 0 1392.89010554 697.217725186 1746.43558957 1270.37574303 527.57159444 0 967.207854301 0 0 732.091646383 586.213630299 626.408868361 583.075384849 0 992.329314296 1592.43777968 389.596545991 1279.32485571 658.977209654 1359.52210585 1413.83124966 571.132799903 1441.1479449 233.970657069 480.597500864 465.934058514 715.963844389 314.043354124 1433.46833148 609.090384844 680.298380197 1188.12855443 725.079362465 1574.90509871 592.431879228 1165.99115112 349.388581175 520.32592422 483.519588986 1774.92291235 518.746829085 561.386917382 514.66366908 966.706497346 998.710394898 479.213943066 1054.96249478 665.130381097 637.089005688 425.624833227 619.790349646 813.873365667 0 1547.17182046 1449.77526204 836.965928908 977.568314531 543.409523461 889.497363494 863.783066931 1378.67125079 408.561654667 465.57670406 0 1547.88961612 695.892495108 1520.96165146 1240.25642502 249.346697942 789.242535588 712.745529506 682.66694007 704.014977 535.633573141 1638.97595043 224.154830587 448.132539315 41816.056141 1536.63284244 741.430162862 1405.617096 1432.31235837 1622.51798085 545.95564838 748.121470797 259.15031646 413.293520048 638.825727506 240.637089713 1026.52472577 845.414677584 577.996912404 397.135650203 1624.47371789 570.919432062 676.12164695 714.005485799 486.364272433 0 1879.03401787 1752.22244202 623.93780729 1723.32543866 0 640.299414125 518.837493402 626.983371322 603.182411597 613.08341788 620.767785783 1628.5129399 1551.39414044 470.707599496 232.521280866 950.40282582 403.415480433 1163.6234728 1525.26723647 577.60781762 445.213009659 540.209902618 0 506.952567351 753.20267726 585.553960488 1433.93392126 571.426082048 0 1597.53074541 893.178912959 0 353.89814667 564.409023789 1645.44051871 527.433033603 1277.86370764 1110.91612151 242.783647003 495.366299454 0 1428.12152066 555.810373683 744.105939524 705.448772264 1720.28198588 330.226298485 631.132814223 0 1415.80916702 2046.49913221 1213.55471769 852.865762408 1297.61118158 531.744891248 817.077717155 1119.81477782 1043.19880302 1741.17011312 435.343934768 493.255676029 902.47995168 607.268541592 1906.2709114 1144.4910527 514.037263428 2007.41288885 1192.40367894 1488.11092878 538.496680119 407.601959628 520.955181535 603.051385515 852.755425684 588.224362456 1085.95403554 726.626351252 1534.32185575 308.602920545 1761.34255114 718.489363392 723954.977973 702.115627125 599.326342011 674.701907784 1594.83521954 278.465420866 260.730304762 342.29832413 1471.61471672 676.082386801 361.804089289 773.269367537 985.680649131 557.869543307 0 734.360884708 0 816.089258991 686.871984689 543.957956034 363.165826279 964.282543844 445.529295653 820.260110252 1481.39865943 609.755757026 626.820964122 385.356015018 0 593.710574562 404.283880541 759.563492759 832.311751705 947.095971687 626.035387185 965.706982443 308.224289742 1432.54913456 1579.92575105 1541.75141658 613.703718121 472.049465902 0 633.299320484 1466.60787175 666.462915073 735.729548471 733.025547017 1687.11497796 493.260928734 1186.50650538 737.459086797 555.152078694 894.77142002 0 798.994063249 1472.17979789 978.184055699 682.454954129 434.179069848 1490.79341952 591.978498023 670.108645679 509.911909496 552.440655849 683.373927521 637.90888906 592.53122362 682.164193387 569.458692354 510.704993914 1133.7968468 373.291039641 1888.14975773 2020.20069634 360.326285146 519.694128964 1936.37146558 1365.33125547 1240.00479077 544.344809091 451.41200771 721.935831188 836.454265096 586.506626016 406.811088017 0 439.320732651 773.375110296 680.39097759 518.696505279 451.75411861 557.007896564 1171.94167783 657.02869161 1616.35484294 1586.13819895 1830.54823453 1459.84579376 610.462459516 0 907.856267958 701.143169568 314.356679783 1334.91481783 432.696113281 899.522242707 832.52834034 0 0 0 1046.86762743 1515.61332878 1832.40547923 1786.32693721 907.698762243 451.708770728 461.547769682 687.656116823 981.458058479 764.97560142 1765.44364499 738.14664744 241.74539398 648.978998632 1911.8309443 913.521882475 1290.58839693 1220.76140774 13541.3175774 1467.28782972 716.131086178 1645.67658386 618.52712986 835.101282523 526.043462954 1254.76178112 776.822147112 723.9171495 1599.84639959 934.467186352 740.139519369 0 472.890902731 1530.59716563 1604.45169204 1701.91373419 1334.79909915 711.585473574 1469.46069242 1261.93295095 593.615630373 459.673552724 1012.95652054 951.353267474 578.780292736 532.964375452 1286.9808015 1175.23974843 1803.11359806 564.538303015 759.155049121 1461.16550999 1563.2091449 693.011710779 380.517571985 1960.73001199 682.062488381 0 1354.10771533 631.577686941 1199.337674 710.343508742 352.644758523 1089.7797435 460.150007111 1079.42466338 845.974899353 1002.4591778 729.740654475 1254.04737285 1221.95394589 685.982563791 0 716.71771122 694.341706366 605.321850804 0 498.562470927 346.561425863 0 1085.09178151 972.95119014 618.604706703 1058.71362583 588.438065644 1211.34100565 527.265708438 0 537.613343328 684.252362377 728.909220424 700.172654294 431.021309668 1486.7025627 1884.71902959 425.460846868 424.042741278 533.039241259 723.829354526 502.550999857 725.20867964 1541.18016671 709.797591369 1177.77004147 1334.82408098 523.962972772 2039.6016337 392.275477554 1205.45188935 670.386054301 1659.82812998 1674.62617878 982.874446783 634.656177163 581.705114591 1379.80935285 833.442928772 831.436870721 1411.88013467 393.667206163 414.862542905 584.530410297 674.080175975 710.522943074 1775.26248885 421.666852924 1310.01587453 0 1804.54009552 1848.49787835 1686.22642153 1092.18471306 650.460737408 0 934.315471169 424.540260171 636.002960761 712.890221109 976.661263731 349.166372432 481.945571326 518.823596431 513.462364933 857.817986485 1229.26857428 1013.53244244 538.769178233 740.500086102 353.310027909 1456.78620683 540.522186185 1576.71471055 1977.67126315 0 472.586155887 0 594.200291912 556.314806599 984.75780598 1754.221544 477.379689681 567.650854351 726.581688318 241.912820137 777.003660917 1579.16609283 930.286744185 398.79809389 409.861150649 515.465442395 923.651558519 1720.45822252 242.515160652 565.551329129 817.513165607 643.50905786 536.652633823 753.429596076 442.204772755 0 685.296013487 498.823572158 572.565453907 674.975702833 1946.32165045 780.921998655 745.339019702 793.556163147 387.008707443 1106.00038931 322.553979148 719.65674365 1252.5647538 1277.66955504 0 1261.37526874 929.43770628 1004.83276649 21 | conv5_4/sep_param_0 162.72600108 22 | conv5_5/dw_param_0 1529.79769174 832.005273086 1319.84028228 906.61723142 1381.20849188 1849.27738882 1404.74946688 1358.33959599 827.418321278 0 1923.45076515 1616.19806198 0 805.905606411 1569.65690205 1561.94360632 1161.80321435 1476.12696891 1799.05530841 1614.25683819 1137.37363381 1254.29938603 1217.17772772 1863.22855994 1753.77303556 1524.37625087 1063.57883414 1142.05582667 1443.24475556 1596.05600149 966.950528515 1395.33515464 1344.16113483 941.302074795 1425.29486514 1170.99038066 778.556823021 977.968660071 1145.70224535 1771.98584004 617.790223493 1003.14812609 1068.40451251 1232.40243975 1287.24214369 649.651719036 1106.11522107 674.773327422 1469.41116286 464.693971144 1559.55492177 1002.92690816 1025.72713307 1748.52462843 1101.3431839 949.780801309 1376.58385534 1775.14435251 1772.23805576 901191.949642 1431.04398464 1666.00675213 1586.04890966 1227.1725618 2040.40731366 617.028311775 1479.91802861 668.716591674 1658.44140253 935.608111931 1570.94004298 1084.21682694 720.139130717 1210.62684892 975.947853111 1448.94243196 878.066015472 1188.19613598 555.591348113 1488.27931678 650.548965096 484.965684007 2007.06826703 832.850993139 769.019660616 1579.57934996 1344.00205381 1658.07455885 1788.12081255 1379.24441947 991.778961544 2073.57011468 1323.63513204 1003.81402101 1794.67959755 950.050672412 1712.45545536 1337.41586889 590.07599196 787.007537993 1411.54914069 782.861111188 607.52597207 978.387416048 958.535785337 1643.64888797 1878.08167498 1706.05897083 1876.32837277 914.688572706 1253.87201078 762.366682884 656.660768798 1101.88347683 979.353942741 1628.92627959 1057.68893754 1769.59768351 1430.6385025 1049.40753411 976.620974507 0 991.471025956 1747.96752894 1812.54246935 1193.43822791 1587.20942058 1869.66149024 1651.40573445 0 936.444697157 738.063356626 914.596403737 1910.76517879 591.81625115 1363.16759219 1119.8403059 614.117093909 1060.00904788 1135.29392039 1967.93416903 1667.52067358 2063.80147098 1628.6945283 1237.17367433 1429.1579723 1384.90959238 1636.86333237 1808.12419307 1950.38660757 906.751112104 1179.05061423 1463.92010333 1358.528616 359.566899627 1122.89387751 1867.5016791 672.792234699 1549.70202728 1976.86987764 1568.87935912 619.520801656 986.340670494 1212.88394678 1463.65663146 1436.44276173 1513.23355896 1828.18186738 1812.36825277 1027.95145915 0 1469.47386708 0 1674.21892309 963.393108484 1440.63467788 1312.59177153 1480.35488739 1166.03310559 1667.29705491 1524.05677554 1088.18170389 1922.41883896 848.513982797 1012.08743389 1645.96627718 1211.87099446 1184.10841059 1037.40734249 1100.60049722 1684.21178304 1743.38242703 1181.03107632 1144.94176778 1698.91946632 1693.69736707 1225.27830121 1111.17494558 1235.65702806 1025.40917081 1880.12335609 1068.52198473 828.166850999 1929.79958091 0 1742.82164934 520.194222509 1329.92500555 1583.51728397 0 1408.97920287 648.950733254 815.808586697 1474.56130394 887.892199786 1677.79145293 1492.09029405 490.452509266 389.226393551 0 910.291829956 1169.53671622 1498.24484455 1501.62912878 1010.65528847 839.10394193 886.529564488 1789.34690715 1271.02300997 1763.63364547 1013.01659976 1242.20202454 1895.25762911 0 1066.70519817 1516.76114563 1623.16674213 1883.36733165 1232.80014488 1933.74024344 737.073745118 1904.14229862 1550.17471934 1790.0914173 1256.65646434 1211.36390426 1068.66802383 840.004479327 654.782392992 1347.87070709 1498.13910514 1066.41710036 563.185790221 988.204349487 1233.06786237 1188.74203931 1269.50860805 1072.97604016 1042.15134467 1250.27855728 1314.75075114 1725.39095956 601.614823392 1252.71443223 544.354335357 1441.20497007 1347.16113976 1478.84849917 1405.97917581 653.897870925 1391.27357155 735.171705136 1009.02394951 1701.49038233 854.828251332 1367.01234218 1070.97752139 990.247291288 1022.83222646 1199.51195599 1009.86169458 1928.25965589 993.620405461 0 1127.90842726 1369.91195656 1297.65583235 1513.90003366 0 1590.47897536 1861.36423513 1463.89583884 1052.65796359 1449.10047038 1912.22257353 1078.52087739 850.882700581 1923.87409484 596.534466896 244.492866388 887.248503993 1597.37100829 1542.75442741 572.550299065 494.185889826 689.930366688 1620.12858035 440.269581291 1694.56248218 1299.35177105 472.430184887 1308.73196482 1374.8104077 0 1274.81018663 1021.50823123 1138.29311814 1830.56946587 1326.98352758 1070.01298881 1039.86406797 1164.16563016 1427.23497804 1605.92096834 1635.2222384 1286.39542654 1223.25783603 1487.32328565 1068.13999324 601.539750783 1698.8789976 1747.43120577 673.358760369 1608.9365535 1356.29837519 0 1371.32583645 1253.4131283 1471.7003534 1090.89027554 228.974207031 1593.65325375 1419.09821366 624.278927932 1175.05195372 1414.12131428 1451.73590199 1216.12227965 860.173460053 1313.91162681 1079.08887079 1017.34259357 702.200142585 1752.56545996 1876.86222387 0 1144.73224067 1680.64436627 1270.58899452 1092.28969427 0 1688.60244893 1192.02827247 602.240420309 1074.79190415 990.24809667 1163.9874779 2015.080365 914.66157774 1721.65046044 0 1806.00061494 1641.45055118 678.637439512 1442.82671293 0 662.202353691 1366.37700015 601.673603715 1247.22350313 980.677815839 1884.64296983 1108.3504238 774.024514638 1255.06675316 1644.22187398 1646.55074143 1269.8885257 1707.67670332 803.993288693 1137.69709914 892.094614414 1386.40073429 940.870103288 1533.02556381 749.832969201 1824.68898308 1355.25280172 0 1260.43024068 0 703.07651787 1411.12717607 1658.20053216 1520.10618809 765.689726261 1646.27928599 1828.49760481 1556.46088803 1536.59890469 933.807926803 305.913007947 1010.91578032 444.519732479 1341.63386546 1974.61714291 1060.74817668 1822.0151579 1615.09501891 1772.60278088 1990.44732219 1961.8265197 1364.48106602 1055.1722546 849.066394205 1852.24279677 607.220819515 1645.98010492 592.251233621 1117.74470467 906.339275579 1295.27348915 1182.15107182 493.236321918 969.476035207 1759.19301444 459.809603101 1790.67738868 501.696474431 1281.8286846 643.774163968 854.397891099 1243.39965342 1490.62850314 1220.48179128 1943.72871898 1231.09224049 1109.27228701 618.078571627 959.413786303 1608.75023392 1097.11419187 0 994.752753192 937.145811786 1539.81537158 696.428947555 1497.34699543 703.82143529 1425.21263711 973.562683789 912.843925024 1620.69036736 0 0 1453.03690206 965.727554335 785.639159792 1363.25927969 1628.00853275 1054.35236699 0 625.536711858 1314.73624992 1222.23748002 1666.30511388 1502.62642855 0 1892.53199067 2041.090687 1044.91904047 1878.54302663 1580.19158439 1547.92138367 1719.00079861 1281.97088033 1613.09995831 1584.73921052 1430.45397374 980.62252659 0 1459.37384818 451.777777825 1301.70372385 1057.83977676 0 1652.96775706 1168.09298007 1302.05243255 1752.43609183 1275.40472114 1115.3914178 862.434897685 1094.43115742 0 1618.14073442 1319.53029582 1197.29303802 1394.62221304 940.019997258 1528.13480063 738.447878533 23 | conv5_5/sep_param_0 204.163206657 24 | conv5_6/dw_param_0 1864.4003763 1597.50723944 1323.85923778 1607.37550704 1008.71619723 1771.03565583 1529.37397978 1363.17576834 1450.5588081 752.491911452 2067.58061146 2139.45385652 997.108959926 1708.08602332 932.5734134 0 1808.50057807 1026.82692643 1729.84053763 1965.47304413 1483.06430145 1853.49131964 585104.060227 926.663538366 1444.93135469 699.976619991 1272.66549328 1666.45254272 1801.53816161 1828.21108322 1579.25475514 2083.23166557 1699.27479339 2139.17234023 1138.78141457 1829.15884937 1692.00958845 0 1261.89717053 1360.19699847 1941.31446236 1073.2103905 0 2004.0497371 1150.33058043 1105.37640866 1822.19318208 1818.31085388 0 0 1072.62123042 1058.25529928 2335.73074555 1789.33037786 1925.16564104 1982.77643884 1785.74867321 2289.37330932 1686.06363215 1061.63107111 1890.68786197 0 1892.58872549 1275.60391352 1301.08402651 1253.47783278 1699.71721217 861.590246069 1255.85236191 1637.60684085 1677.8842689 0 1393.28494657 1430.20180941 1705.90052453 1350.54769814 1349.84610648 1492.09708581 955.472236703 1521.17678772 2018.93945782 1690.50303421 567.534506444 714.005126901 1856.57197615 1864.84972544 1334.98361044 0 1839.93068611 1720.7043197 1860.01456048 1864.4140392 1535.80380521 1425.20274655 1487.7692027 0 1922.73023079 1655.61281373 1845.89578738 1755.48093921 1939.83668622 1998.22065585 1455.68618706 0 1738.91816961 1877.92628631 1576.98281992 1290.94613352 1763.75682463 1673.74349391 1175.94739353 1510.26997102 1228.23179829 1490.58418401 1723.6514832 2793.33326822 1775.88393404 1963.44655569 1346.45316998 1482.03776761 1137.66475192 1564.929894 1896.92659069 1325.59182991 1990.42477696 1117.91411403 1359.86689977 1822.21422006 1872.55341006 1559.63383228 2104.00866604 1988.29203272 1158.79945595 2414.91438553 1941.7091965 1627.34517598 1660.77256596 1854.8273021 1534.20778689 1504.67919744 1029.46472453 2092.00922977 0 1924.1512053 1602.77423609 2339.81175235 1759.12765629 1550.06998064 1605.61736958 1548.09992469 1420.78554027 1454.77031689 1980.23275309 1956.86918132 1651.89480996 1820.91428474 2009.4951893 1492.89711656 2160.3682886 0 1559.15834974 1248.46467097 2066.91021477 1474.38031954 1610.47568072 1366.46353321 1118.17708978 1490.86096103 1504.62686689 0 1313.24797812 1311.89017763 1905.38149444 1828.41405125 1786.61883134 1579.90920351 1619.40054696 1440.47178537 710.579990199 1505.83897388 1926.92648895 0 1408.90431984 1635.42242906 0 1880.23078314 1716.9702557 1418.51423249 1203.25827035 1437.29521379 1919.781913 1292.69250209 3161.36143403 1706.63785728 2328.72489609 1700.14629536 1799.55672694 867.380633802 1849.70582782 1978.79530577 1956.17525329 1242.68688717 1422.03768469 1854.64768764 0 1341.97947011 1760.92113741 1235.46438429 1201.24292044 1957.24734344 1187.98455519 1471.26845994 678.641276168 1339.99425 1575.22019258 1650.35894114 947.124704636 1956.83031732 1229.63560721 1796.25233109 1671.88333731 1858.16860974 1157.55689085 1663.07573013 1240.55312188 1654.29107442 1804.41822159 1742.01533978 1347.72193539 1372.73346215 2097.84590486 1591.45844342 768.133889295 1993.03760833 1714.10154957 2192.38366547 1505.41499963 1218.17302155 2030.34374878 1557.48128545 1242.27716524 1821.86968695 1224.62653594 1813.54660153 1690.29835139 1694.23219288 1563.76700478 1424.02889085 1750.75046818 0 1854.63214959 1601.4989209 1248.98591686 1775.79235452 1161.26420398 1953.85585577 1624.75119295 1667.40828588 1567.44044009 1986.98877572 1150.20196127 1496.9256856 1963.24212421 3983.9808851 1573.07887584 1552.64857477 2021.19481261 2032.02579817 705.657348868 1514.98304573 1916.05293747 1969.85588024 1127.33262502 2071.54001246 925.421502988 1950.56247815 1745.55228423 742.40002007 1946.31342772 1267.15558314 1272.07113595 1390.50363153 624.451183741 2931.04869688 964.711934612 0 1040.51590622 1849.01981896 1855.03561772 1966.73596682 1132.61429455 1925.48118596 1650.29087441 1856.76915517 1486.5349202 1890.39955009 1501.89295212 1461.90085955 1405.7553901 1298.75291387 855.469367287 2155.03543517 1591.46230666 1288.72463521 1562.73491867 761.45656656 0 1475.6412458 1733.90101691 1717.27763676 1678.15732628 1449.33851691 1357.9920058 0 1695.95005849 1938.06495572 2167.88539555 1398.24262557 1497.72261339 1437.74667914 1553.27053749 1779.44824243 0 1369.55984985 1347.7840617 1211.23117323 1480.88282924 1107.00029686 1428.27947759 1848.38262194 1583.87586382 1734.85254836 1898.42235449 1346.33161412 1858.93258009 1842.33754203 0 1710.93839111 1660.70784428 1825.41432816 1012.27075013 1410.20069802 1715.44570306 1296.80296127 1561.15123621 608.566570997 1776.13911144 1919.88310797 961.765971589 1554.97613082 1655.90183369 1270.82837219 1813.71158842 1750.88066661 2006.62997886 1797.45643599 1870.19422133 1650.56717072 0 0 1795.3595291 1481.90776322 1083.03945334 918.238842437 1624.92605711 2267.77726821 1849.23545862 1397.71968891 1403.68209796 1273.03665451 1942.28267616 1981.06012198 1784.68052518 1860.77192458 1665.49333809 1496.38834696 2157.14125432 1065.2098787 2228.72127555 1617.77860352 1889.96735188 0 1495.18746592 1664.24449566 1801.39518046 2071.75503105 1311.65607591 1815.77442946 1591.885296 1466.8532202 1339.9607528 1844.88727147 1677.65555057 1823.29150702 1807.38914045 1593.75845174 1959.26352622 740.507871065 920.134585309 767.006277917 1285.57784278 1778.76954091 1642.45568391 1787.68217704 1386.29621151 1691.20697816 1239.48136875 0 1499.31242339 671.89268922 2238.92383609 1192.67316337 964.605915737 1251.08278974 1867.76033097 1923.60335946 0 2060.50888647 1388.28529545 973.328420492 2002.01420606 1538.61046679 1869.40846215 1177.39452005 1379.81426734 0 1879.89381857 1773.45076149 1237.69218301 1978.99012273 1464.96814115 1749.18492904 1872.90565136 1360.72351291 1739.76156594 2049.09987679 1103.8196045 1878.42958225 1590.26085308 1802.56787696 1248.82020167 920.798349249 1530.78369232 1577.42515161 1802.40262412 0 1909.30316022 0 1524.16797701 1381.35500974 0 1354.35129855 1829.3722371 1363.89401672 1222.21145178 0 774.421816189 1254.45298748 1807.00727949 1946.77289538 1641.77971327 1714.51067915 2124.65912098 1648.35301121 2100.07331877 1711.78769141 1485.45800955 1519.41324598 1644.06582493 1156.50810023 1329.1248553 1471.36713795 1254.46905134 1990.66815294 1811.75779558 1739.53697081 1442.23414777 0 2436.79917268 1907.6448091 1348.32191207 694.113703248 1834.67477674 1925.18738445 1013.98993604 1810.26507512 1875.24569576 1365.83285986 1840.85148044 1529.39428844 2008.41599074 1317.94789755 1751.74237042 1607.57378864 1723.87129742 1459.89880671 2102.56322895 1233.09810159 2147.53505356 1441.2572471 1566.38939458 699.378328167 927.384789751 1551.02061864 1512.46473137 0 1651.37917646 1604.23168342 1236.39986422 2079.61179981 1556.35288243 1557.24494623 25 | conv5_6/sep_param_0 286.221061478 26 | conv6/dw_param_0 2794.60250918 4115.31630586 2688.75480251 3606.92008221 3688.89577812 3241.31077744 2728.20672374 1060.38359573 3531.2409566 0 4727.91342189 3914.92937974 0 3493.21476517 2827.56962568 3662.16895371 3817.65956871 3894.4251204 2735.10431284 4399.33333897 0 1333.67087513 3638.14729782 3842.63873261 3617.7277497 3042.85322431 2006.2733465 2612.44605662 1174.03640786 3579.93207231 2662.00087923 3343.48595961 2867.28991424 3531.09392216 4011.07137069 3865.48084188 2742.20162721 3808.75133726 3493.31212683 2503.11228067 4399.86222879 3594.21625432 3390.75814634 4866.85198904 3222.66831389 0 3715.46843433 3674.95563456 2564.52770953 3553.05125109 3288.12657559 0 2937.76855767 4193.2613596 3620.76279005 3328.88160805 3932.09998202 3232.31214997 2906.79015338 3908.38639704 3549.63920968 0 3807.75715279 3693.03453706 2305.61515282 5038.76112285 2496.22617918 2566.661026 3874.79501958 0 4695.03386754 0 3793.11405455 2653.47128823 0 3336.63978978 5525.97085352 3984.12615021 3564.34355195 3442.73447817 3891.91716709 3960.36842821 3347.13926375 2958.47291634 3792.25625195 3220.64494441 3926.31828521 4162.2193722 3684.35691052 0 3611.94765345 4048.33819776 3936.52096345 3385.37881199 3110.90863458 3891.95315633 3443.62960568 4025.66632174 2693.29800595 3387.74750649 2664.21873048 4140.83812778 2942.69084006 0 3805.2401939 2591.90517989 0 3550.29499043 3756.87572479 4041.60272196 3544.59846423 2536.78554988 3392.83079279 3740.94418437 3467.52585385 3455.15958163 3490.07275659 3362.42585977 1192.42953749 2858.4777173 0 2922.24685296 3482.67498363 0 3750.12709816 680.700669437 3322.9848984 4096.00812009 3634.76486262 3771.66457609 3252.90404452 3850.13082062 3655.5937091 3620.36981923 4228.05247298 0 3762.05009129 4515.52128429 3331.18489519 282411.416035 0 3279.61375265 4097.04332671 4373.7874926 3760.5611209 4153.63689 3731.26086208 4697.94341737 2303.36649667 3552.91979775 4370.56023043 2800.11989476 3254.91253528 2633.65646566 3596.05654696 3817.68521975 0 2947.86795855 4026.26347527 0 3430.57964739 0 3817.39623798 3753.61905537 5983.13884845 3546.08764275 0 3502.36319957 2540.74447989 3950.81646453 2654.49711092 3699.16157481 4283.09388525 0 4315.03171782 4149.29249745 3975.10493713 3700.02274977 3976.71812235 3338.74422103 2145.01353824 3186.35232607 3865.07502619 3829.38327853 4220.44735552 1159.93220294 3405.48335907 3959.87897009 0 4047.89788811 1040.51368316 3634.67728221 3650.63513826 3931.25614482 2771.62698302 3310.28839016 3627.85203952 3097.31126698 3467.68351441 5055.1718388 3725.725695 4543.23240991 2587.25861489 4555.84974502 3764.26996458 4055.13667834 4329.32564244 3135.37505176 4083.62778934 1078.6229069 3968.55253737 4258.69874206 3997.64804593 4463.4755165 3153.80195417 3270.33686751 3265.51865853 3922.18188219 3256.70976097 4118.1678021 3997.19619679 3620.62166449 3296.95629858 3893.30567871 3914.52660002 4035.13175427 3448.75685231 2561.70284977 3318.53688896 0 2931.10640627 3670.89838407 3807.79755667 3606.17531668 0 3587.01012321 3897.92234753 782.511356241 3758.8872302 3558.5116274 3012.88286626 3889.23184982 4253.45252759 3952.58595931 3982.90476338 3651.81650436 4132.86876346 0 3906.47494526 2596.66517697 4325.4550102 697.041640809 3823.80813539 3840.30472898 3774.3324328 3812.98572461 4011.36351637 2620.07020905 2609.88768281 3852.87342986 2285.36049007 3400.78544388 4010.49900183 3380.22497339 3780.62761476 3748.20284028 5012.94071729 4128.92170191 4365.41617601 1103.50618364 3930.20740788 1171.3549854 3138.53866902 3307.98982548 4166.56950282 2842.76506933 3806.07668577 4809.28817525 3411.7713434 2582.22276967 3627.94083564 4743.61463284 0 4168.43717544 3759.21840709 3538.04350579 3986.58282713 3996.76787814 3051.33856919 4705.12951475 868.342568606 2820.90572122 4180.43060738 2616.84611184 0 3120.08670672 4065.40883249 3548.53890014 3712.76706473 4367.75710305 3726.23065699 4470.74780622 4611.47481705 3480.92755485 0 0 2852.02225346 2705.58895952 3718.19117534 3596.47612739 5529.42781725 2869.00097798 3362.87097436 4058.42901491 4267.35675855 3650.93109235 3341.50032943 3939.65393988 3443.1339939 4049.97433514 2608.56107216 771.114867612 2912.40081852 3558.57180232 2417.22922321 3707.61397916 4021.69662065 2662.87709335 3589.94959015 3499.2697298 4630.65303323 1106.04258721 2589.86605017 3446.38537244 4099.7716785 3409.65128592 3043.77909455 2431.43219662 2484.26573384 3695.67280377 3501.55272491 0 3659.67532352 4000.28712163 3577.87729599 3820.16430808 1140.52851564 4322.93196848 2773.82595424 3289.67717834 3582.54969975 3953.38350564 4682.5844065 3823.52980434 3535.66795819 3741.2254017 3756.49404703 3476.72973819 2441.01706484 3905.41254278 2331.96331987 4012.34787433 2623.81125166 4117.59803223 1148.10534873 3735.59720559 3671.35972986 3487.11009324 4560.17226493 4159.92270966 2755.38023497 3132.4463197 3491.97459488 3969.02565855 3405.5894998 3455.36024702 4518.92942342 2414.30948197 1122.67496486 1019.37689693 0 3930.7711364 4575.02204717 3546.61555241 2770.64712838 3382.4693297 3018.28635457 0 3801.30016579 3199.11420223 3883.23658783 3466.42791819 3559.98464033 3612.01845121 3840.71920571 0 2643.34544894 3618.68201492 2987.00993088 3450.57864185 3923.5988502 3512.35307968 2953.5849955 3222.51021324 3239.1894556 3443.59029933 3816.63039144 0 2876.08381882 2683.17586408 3742.73648816 0 3340.04251381 3743.94738774 0 3600.29896568 825.339688795 3341.36932609 4130.4822559 3295.0169352 3416.55960005 4041.79151247 1078.5077071 0 3637.37172988 4051.04753566 3887.31293087 2363.67988026 3110.07113772 3785.50781241 0 0 4225.49509928 3434.43555582 3098.99073593 3909.46566126 4159.17412524 4065.08597966 3567.01754552 3037.47136938 1171.72384449 3798.56655132 3472.81016943 4374.04731617 1187.00051566 4046.73316382 2634.1135115 3807.25961737 1861.42562128 2854.2450699 3962.40944712 3689.20076309 2651.3508783 3388.14109592 3217.35532288 2472.26157544 1164.63030677 2131.03101466 4326.51474255 3834.8137967 2913.11531223 3768.27476117 3330.97137946 2621.16810379 3827.0116427 3267.91486645 3847.33523387 3456.24408241 4256.55241321 3937.21245525 3789.89286897 3000.60352267 3957.17487224 4126.2810287 2856.00471621 3980.87140175 3600.46512894 4530.17783916 2844.1828557 3736.76252998 4882.12746534 3656.41942129 4010.06263904 4097.42027577 4851.33798915 0 3259.31054134 4420.31862187 3732.01652359 2613.6161138 3404.16633585 4238.27242634 0 3899.12917576 3343.87031497 4240.83764501 3669.52610156 3668.90766533 3864.72844116 0 3285.06235569 4190.40900246 3208.34409676 0 0 2609.23908664 0 3169.77134744 3429.1724324 3896.5336591 3210.59783699 3956.74039169 3831.64717464 3901.77146839 3717.87286386 3275.71122802 4078.31682992 3948.51524151 3548.7272862 1047.0795188 3530.25693696 4221.71371522 3669.61615947 4140.63745671 3453.16435239 3174.48387156 3952.29727188 988.05106484 3399.89481692 3831.76388515 3700.47939534 4018.99559036 3391.14299055 2824.84008544 1143.26501165 6007.31086143 2731.40746251 3675.49210112 4354.51901226 3808.6568738 3831.38449435 0 3561.39488019 3057.90731352 4003.35979458 3719.27992867 3845.50730861 5095.30393724 0 3366.24101444 3565.91986516 1032.79900409 3951.8624874 3523.51267372 3447.7150494 0 3959.58323779 3313.27170812 0 2305.53142139 2604.56922171 3551.42857625 3668.32654111 2597.73008002 2564.12631145 1025.36765837 3118.90752495 3828.35937903 4912.76046337 0 4754.79937606 3385.86533332 3111.54862601 3712.93568415 5524.06899329 3690.74162797 3508.94829054 2928.95153316 3985.4004674 3785.89877256 2309.37075622 3587.96146662 3945.34396171 3050.6532177 3765.68367323 4315.42335413 0 3763.77126198 3659.3437769 4304.0017685 3652.36580072 4336.92965156 2637.86968803 4875.77585904 2592.02046409 4492.81946159 3548.14409307 3717.78163813 3482.73368842 3535.88505214 3589.29835237 3902.25693978 3982.11805448 2398.8632055 1424.53098669 2584.80555688 3842.94151242 3413.03856166 4032.70028376 0 4021.59651775 3776.29491415 2654.32143546 3064.2548435 3857.2218216 2444.07539136 3741.04722406 3645.15323408 3732.20037942 3500.14670609 4013.21602084 3890.15983644 3450.94993783 3951.679256 4491.74801796 3959.30180422 3576.41570525 2820.36196312 4292.63102312 4107.57617772 0 0 3341.44432425 1119.8364803 0 4510.53029809 4724.66509544 2446.05207732 4005.99370957 0 5289.16835731 3100.33393131 3999.6033296 2638.57589157 3489.66013115 4029.54003441 3856.3286747 3911.87778242 3368.44719649 3994.11144867 2654.64014859 0 3931.38444276 0 3007.86407833 4280.91080739 3548.72137575 0 3831.15930698 2860.98405631 0 3859.06614531 2935.62966566 3998.99201006 3738.1814724 4412.5552931 4494.91498877 2046.83887227 2847.76705771 3440.26642246 3407.82270963 2111.68817238 3006.90396445 4308.74850579 2022.33841677 0 3859.86310317 4710.91774083 3579.72908209 4342.58970764 3804.59215494 4980.92053244 2412.9053662 3012.84505651 3325.41755972 2410.85075537 1308.55333239 3543.79706014 3932.50185019 1133.32585054 2607.9364768 3980.1956248 2621.13102225 3834.53213601 3601.38253492 0 3588.74406043 2976.87962916 3831.07664445 2647.4281488 3202.87145496 3725.941916 2351.86298593 3973.2531742 3130.99723199 2516.56617607 3367.74375146 3377.07041096 4373.07916722 4436.39538914 4079.74000185 2472.2798627 0 2448.73197753 4074.744246 3697.11322906 3910.21181344 817.909437926 3590.54433921 3782.03391897 2725.71045554 2922.81958507 3750.87556459 2662.6364617 3145.99545606 4407.17351733 2422.71627296 3666.81774138 3227.10815984 3847.40557339 3625.54409393 3652.76653135 2954.62453539 0 3004.25469678 3610.59308074 2621.49321683 4511.32892623 4027.19426589 2457.44811962 5803.79099762 3906.67863095 0 3524.83767001 3373.01174414 3734.69893627 3456.15928747 2662.81344756 2832.15114818 3588.55442384 3757.92636002 3578.32419407 1004.7209444 4376.0506211 0 3628.88272813 4541.60218537 4230.22290177 3681.41829637 3996.56124956 3083.62973689 0 4439.76805892 3977.61175676 3879.4998643 1170.84962078 4294.54582183 4393.8404696 3998.30068887 3899.34235421 895.595825932 3704.44976756 3722.34800888 3772.66337996 4413.55700094 4312.05525797 3441.23078016 0 3872.13854752 3115.11598745 4067.50353083 3142.17691098 2571.52844572 3758.40776956 3856.75316241 0 3129.09448366 3481.95076491 3568.36501147 3333.63153759 3701.48786873 5544.61621703 2431.64568632 2892.69266319 3959.4071072 3980.05808071 4325.07060475 0 2322.00274121 2618.14999882 2892.35398345 2757.63044651 0 3695.20572827 3809.23266304 4673.56491459 3915.80895029 2605.86249353 2513.15173762 0 4178.99985266 2758.56004235 3519.77585305 0 4069.69099309 4342.35905046 3680.86658605 3623.57759947 4449.7870915 3771.31743511 4012.55047122 879.284339732 3602.340757 1136.14294552 3865.55666787 3445.84647456 3322.75559187 4136.30363895 3795.77938736 3336.22313968 3642.82245141 3603.5946742 4153.1154469 2928.96788989 3054.10822884 3503.2345274 3025.74777553 4015.61978846 3234.76083203 3542.17141584 2581.12187911 3686.89307283 3412.55137395 0 3551.52514017 3307.74942519 3814.55662783 1189.07796761 3662.85871189 2636.81771948 4013.2854699 3512.0657783 2975.48984056 3599.13169246 3451.63161189 3637.11405678 3760.93864675 4089.45829908 3747.46367875 3176.59407512 0 3920.34978349 4072.47594231 3498.43699134 3575.05240592 3484.87509689 4314.18587306 4074.36853579 0 3414.76225859 3810.85883482 4894.52274621 3223.99800874 3975.55922307 3780.92028142 5761.68118248 3750.28386353 4163.50670279 3847.61747589 3941.51962401 5721.70977959 3720.39530412 3705.0391728 3077.63652254 3706.79601995 0 3743.16428273 1684.53601241 4241.21117992 2826.86201635 3563.95080911 3655.99985329 4019.09508999 3754.95363564 4125.31336172 4924.17905012 4080.57577273 2732.36522913 0 3308.97136777 3090.47539123 3808.81048568 0 2891.94914436 4493.78834446 2709.63007468 3754.6422314 3431.00604263 3466.24993056 0 2015.35482734 3597.97922581 2477.1945795 3024.23472771 3967.61309791 2166.84066696 4101.84420265 3780.03990257 1032.13229632 2901.07811584 4065.51064356 6685.83032956 0 3734.94566152 3186.54859707 2953.86957392 4509.20852751 3236.97900773 1420.73213259 3758.87562555 3445.42822206 3458.193761 3495.00250269 4829.53011091 4373.77037788 3608.46362614 0 3644.87107571 4073.14108205 0 3525.98131188 3566.71264945 3713.11402495 4037.67327482 0 3482.15170897 3422.65147586 3709.39546803 3627.70109609 3849.66909943 3989.46593192 3727.00587705 4747.37691507 0 3614.74761851 3221.15861394 3702.87571912 3805.50397414 3411.42959475 4143.79613156 0 3604.99203178 0 3668.13945197 2533.61654492 3655.85557552 3715.36274973 4469.71733586 3884.69106343 3842.71149937 3848.70340817 2079.20111181 3608.85745463 3981.22146447 3848.41970352 3516.94614388 2850.68674441 2518.93658921 3921.97386954 5644.66992104 0 3658.05940466 3848.73816812 3926.99895925 4662.64736641 0 3103.24272484 3393.9954473 4097.51286172 3381.71238172 3915.3758608 4833.49061479 4103.93981969 4455.12911762 3841.88177531 3540.10570607 3563.3782453 4098.81270771 2671.25677805 3364.20968143 2918.38914774 3436.16327492 2328.17050435 3608.56369856 4190.87158955 3454.0901097 3784.6404215 3819.4585418 5655.79692666 0 4326.86837666 2829.19438694 4165.87655656 4368.95244393 3256.52777156 0 4047.06319706 3844.77436999 3496.87994754 3755.43924809 4163.38492509 3771.16558103 2584.51103223 2318.43236383 27 | conv6/sep_param_0 317.761344308 28 | fc7_param_0 211.145021716 29 | conv1 49.4665182589 30 | conv2_1/dw 0 123.720796452 48.7053492114 66.2342454279 49.5107042956 40.7008407221 117.291285042 127.183699571 0 0 0 97.9339637418 144.1077263 0 0 53.5370949826 53.9831217202 51.5210828331 76.225973214 172.926695041 135.222714812 84.0615753965 0 54.114960539 105.11281593 0 0 0 52.9767664258 0 66.6691719992 44.4904826582 31 | conv2_1/sep 45.3043611898 32 | conv2_2/dw 36.8117911543 59.4094298574 0 58.9557517643 118.804872994 29.1171263542 54.4988422125 31.9838166694 85.0906895352 128.89308426 41.8878923968 24.4024948061 27.2154362872 66.9972332767 41.5182622905 44.7363733899 162.215163472 124.776862867 57.1518606663 37.3941705234 0 0 0 113.889215342 0 122.059173629 126.916295794 0 0 36.9904510483 119.208113325 30.1652319146 48.8089420511 0 26.1061315041 51.1389507487 47.1234068096 106.953515276 88.7127930938 47.39689769 55.393858704 160.614887052 39.9148013479 144.492822551 64.1094893864 98.9972526324 0 99.4301780685 132.54023619 0 82.4123539168 89.5660581684 141.218484768 0 124.741639969 31.0432342607 29.8571496102 97.696203511 41.5538745258 96.4501358205 24.2930302202 102.958603899 29.3090973399 23.5554622903 33 | conv2_2/sep 36.1855828174 34 | conv3_1/dw 119.401036742 204.568073915 148.767954398 122.405735941 87.6567958436 124.334345746 85.6242036928 186.419611751 167.207559557 114.730470607 150.364335476 118.710841723 327.803102172 199.840057527 126.544060947 72.6922595397 205.588094572 86.415230802 157.01010146 90.013970444 76.2843378709 134.364880709 147.981254217 108.048656753 78.3266712164 109.537417605 200.157532108 153.026097408 66.2632721918 124.192770758 0 145.105042167 123.369492187 123.000682812 109.541588252 109.402525877 106.407712177 62.7005640964 145.633291009 142.678216161 102.013349289 212.685429049 164.739998624 82.4765283618 118.729416139 82.2309568956 154.916539481 116.961548906 75.7357321524 184.32101116 266.768210117 95.4489716475 89.9761394307 162.015118419 66.6619118779 148.140960133 63.7817558336 77.6865073189 154.999756582 159.808556766 143.902540834 188.648066442 0 229.296880441 117.032194027 176.089444459 144.597815764 44.8486051074 250.667304148 171.25859976 79.7897585549 162.653190937 199.45810424 202.968250896 134.468345455 121.749882214 0 74.4518047577 138.519213674 168.382953283 135.824734067 176.467873597 84.2517843552 159.239933836 157.269475958 137.269139674 0 143.001741151 104.153844579 159.261867228 138.897198163 156.018647582 89.7201288979 45.1597294543 89.6061716658 155.049354196 51.0438268583 130.797591345 92.3887990636 152.258880816 144.208077291 102.316924538 168.592330767 219.731360859 147.979027413 221.384520897 68.1577331034 0 195.507498974 70.4363577448 168.155845354 133.22293798 149.353046204 124.396502437 44.7850116206 110.902455457 0 195.748393802 144.709482669 255.348979663 104.460291873 119.410439942 63.0598450033 211.614779144 87.0474240915 113.118404353 90.2064667654 110.219532833 35 | conv3_1/sep 67.197295372 36 | conv3_2/dw 92.7216632611 131.499996238 105.432951937 68.1698975193 75.2995648723 62.4528224416 82.553221338 59.1752966806 161.941290916 73.9813261298 100.423483145 94.1514506596 91.4541593512 134.940077459 102.288396239 80.1502516998 63.1534365005 63.6180918916 85.8396138704 68.3403227752 104.97941956 64.9458440212 107.141189265 72.8586355812 104.573946622 0 62.32683009 319.88299939 63.8186423847 67.2064835051 81.1789823627 271.81193665 187.732231825 29.2020341321 53.6071006037 119.528689044 77.9123130131 81.5198518645 0 67.9869825492 147.882428177 124.253809119 76.7603130889 140.165988861 71.6933801816 0 94.2329731063 117.207467649 76.5993433869 55.0023652963 60.1898603628 0 88.3720883851 41.0096068985 118.768797341 0 67.2830931475 137.129839288 111.896111826 101.303901689 66.5681246473 48.5908653237 60.6681686314 90.0851672709 141.873931544 86.4704558521 84.3358805549 0 114.477655485 74.4166034829 71.2981327873 104.075800117 55.423104515 79.5036833557 60.9805527353 67.313387606 45.3116772862 154.921464125 137.9911905 122.129872842 58.0405607622 59.0772995544 79.1216515943 122.72637493 140.952245186 78.2565803556 0 84.4148469552 57.8856333044 50.7617012753 83.0442370228 66.5872151809 126.688542784 99.6311155711 88.2147940979 0 68.9636166381 0 81.6551673424 72.1485158717 132.610513538 31.4027926896 85.6064809886 106.694275051 112.258524735 56.0569688399 82.7324099908 45.4839142892 84.3534124049 69.2268653517 115.528607509 179.14701573 0 144.247194538 36.88694646 98.4292568594 83.5608360825 87.8880757496 0 114.187750407 82.0072796066 99.3749144268 125.231134598 113.959349578 52.5062153623 169.046865796 82.8465756939 67.0141106379 37 | conv3_2/sep 51.3101458647 38 | conv4_1/dw 143.756748358 196.289213009 139.161278774 248.110154153 124.038592021 136.280281738 120.453775301 138.959891247 154.360782763 181.420409485 260.838503178 0 132.564206635 90.6282943008 162.94479701 162.704469665 165.49855749 180.595255831 206.959882519 176.819344624 159.108450333 239.823534402 143.140919031 183.839686251 234.835246637 137.49707968 173.100246496 0 133.917966883 166.767680083 183.379324277 178.098469258 0 177.402799199 70.6673633668 67.8586803099 150.12227248 0 185.711331042 126.24530656 223.806878304 155.810596944 170.165572593 139.938860974 112.460510981 155.573055496 164.815965305 166.722267318 144.499227589 0 163.361478763 135.716445239 87.6017192015 84.6310196669 126.1544158 126.808614921 74.5984237461 0 114.728071903 0 243.922167242 0 196.198326659 135.572846824 161.69335564 193.056671033 128.523467049 155.664052839 158.380524459 162.379938974 128.636357482 0 88.8512957859 175.254711638 128.014165009 219.277330757 182.662874115 91.2403755788 141.766168291 188.560809251 0 81.1466314822 159.665541944 197.414804405 0 120.271407543 139.505705186 86.0498811552 171.125666384 129.820813001 240.882599014 86.232162337 152.707523259 123.761965549 173.55132026 171.45730445 128.704095045 86.936704514 173.900761949 201.888511969 257.642691786 127.630130756 127.543084911 162.933200458 241.886420129 132.044710915 148.507674982 198.817914582 147.128050178 159.505103723 126.266170739 0 173.527817549 196.26088792 101.535792519 156.369092486 146.604732724 263.761123865 65.2332300239 163.495944482 180.366558809 163.270398937 168.438596686 140.666471014 91.64810658 133.998633335 0 177.775544831 214.276356478 85.0697142907 179.380215774 161.225672475 112.701695688 232.943111601 95.5790488104 195.815642587 127.330373699 115.184072152 195.686707565 135.487136263 163.643606146 140.702971424 0 106.187551399 79.1981928613 201.17765074 137.393325581 93.12212452 133.778163724 161.291378873 126.574563885 197.661525421 176.438119045 143.597989952 133.058368341 130.898050354 156.074893785 132.678002253 116.039206191 187.240069996 217.391231295 156.25734388 127.749580726 170.435931911 88.088109639 77.0170268446 203.352069275 112.799590089 0 86.2363574351 169.644698691 163.178749085 157.029817451 95.4784707371 135.486602716 161.473126745 138.619225039 161.536023081 171.760647276 0 0 193.734104305 244.656525585 167.977666518 134.71528878 140.110108261 168.245157859 178.438231807 136.143528524 0 128.001716055 160.107061395 178.839871015 139.81899744 73.6890362504 163.825896597 0 105.428176123 121.072896248 181.863615694 79.8653379227 193.654185935 174.663915534 268.414209511 145.646381552 122.800655 96.0638297273 195.051096789 196.70679503 108.255560872 167.492660419 185.200094623 136.315643416 161.733160431 196.368624513 105.646090845 117.785147218 110.936845053 130.420863828 193.47798602 135.719525577 254.428648961 178.778023093 165.781981961 126.650060649 115.67573174 123.690974112 138.419235336 238.43575829 126.430973569 195.246406162 83.3136574216 181.885600518 126.76725077 145.50997086 64.651023301 78.0843841734 83.9713049002 153.006731618 181.857312161 154.372910803 170.313351243 0 200.209723757 166.107424945 155.332426642 164.250790241 176.414507124 181.57752282 104.999272672 195.864569172 204.422252051 132.887556564 146.13652052 136.26225355 134.70935092 39 | conv4_1/sep 58.8726106313 40 | conv4_2/dw 135.354899049 173.194304762 208.326617867 130.479841468 98.7364997818 115.535235029 147.115318319 0 0 118.459831401 0 97.8472063869 148.742317839 90.3610110953 145.787010654 129.083583679 117.82020291 129.369706935 125.211432815 128.031491599 100.056171695 82.5730669244 89.7200460377 143.622984482 93.1908818198 99.6655720656 125.132392929 134.447686236 136.228636314 105.37588157 159.099056936 128.641679292 180.775123417 129.636239485 92.7206255915 159.355421389 106.871665863 123.708726878 162.645999959 113.058069333 93.4209325769 142.103651647 97.2199794638 231.931362436 90.8774386365 110.367317253 167.093103251 72.1620377211 150.303870576 134.37710759 153.347878563 211.318946169 126.320801565 183.323109229 161.678826472 0 196.335496654 228.650504516 66.5039874045 95.1805815406 83.1576349931 221.491472201 0 130.925851036 107.993895604 167.369334054 183.496412273 80.4266878786 154.088427486 91.0596572332 99.0060524549 64.552911436 155.211602884 117.83677265 187.825475671 160.375616426 79.1954146786 109.836779291 152.950471477 162.10783951 101.909220139 148.054484255 161.355468635 0 129.692825048 220.443789419 206.971604946 183.581743982 0 0 144.132126134 98.479018295 97.1941200991 109.346721595 125.714798926 80.4256172783 124.410278985 153.337240473 70.3471332116 128.143252144 105.32486018 162.887609406 142.846363908 195.833790986 112.699561517 152.097528957 148.678517222 111.381905992 124.244593252 162.318535361 175.275491538 140.705180584 101.694024762 92.3306871922 258.231394069 216.941201419 104.071977583 164.026425731 96.2295369274 274.569034776 161.791036614 139.207876904 388.386720481 191.87041345 110.296302398 76.2991765153 117.843566592 145.840751598 177.702446381 95.6206586355 83.8274015698 114.992448653 98.1031152708 103.415627979 101.779836418 234.649091248 137.216505584 0 0 99.4713747949 128.410263198 142.961629855 145.421771365 195.315444202 119.144558328 172.863633975 169.368312247 194.841886733 111.278294234 111.830654029 192.879846621 107.900167267 65.6659314443 258.384873901 197.100137171 111.740726439 112.383989883 0 221.843781897 143.674011346 148.668103515 180.29705764 104.881451754 127.575104832 112.904113789 115.137067387 121.684558548 188.444620284 111.208148756 107.16453082 0 208.327742713 170.328938086 0 197.063623598 117.264210557 0 179.770679598 193.014561386 121.797647089 116.955453276 102.483943134 91.4877225322 140.18678754 128.404775186 144.793701675 128.927619026 112.942346993 161.638139571 138.576076578 0 86.501056246 85.3455000339 98.144767565 121.783318852 0 174.856378293 133.143271273 155.05986123 151.252313774 115.549564927 164.426828013 86.2257125828 115.387808037 142.090705426 0 86.8887959304 108.243499953 98.4457295542 167.96063385 112.133527099 138.948350166 139.49255449 132.313041462 97.438534138 117.517467481 181.697668549 0 136.723246065 83.333318926 96.2879348338 174.485327227 144.194257686 118.759831594 88.3750338846 215.99951907 356.594321243 138.598777028 126.549984409 0 71.8807889673 96.2511905323 222.957945038 122.971776818 132.951360622 171.329863923 0 145.079704849 74.2172793836 79.3481258163 198.111231842 136.301958275 99.1784794441 86.1850866334 69.6227042396 259.919659021 139.236723656 65.0234328341 163.027738288 155.595392124 153.805439725 168.548677247 163.26237542 143.215832301 137.767820507 0 41 | conv4_2/sep 108.638994334 42 | conv5_1/dw 196.947763238 179.045480207 190.977390867 175.51046151 189.606578593 236.614119254 160.117542486 130.40494854 150.685076034 129.476563524 190.409212259 212.462561365 148.846686092 168.738804511 166.585485766 214.80205469 175.638338461 218.925877566 213.647437431 180.637792936 125.247320555 0 189.555995562 117.325874263 237.28569852 204.073615915 110.373700348 171.582898541 239.254578027 130.158628332 184.40095265 106.115082358 138.865714468 162.291587274 190.741804411 160.30962944 185.816772978 215.379556606 228.574811048 167.231411618 159.286051622 212.465294314 147.671345539 0 0 164.393150238 100.051521575 107.398504713 193.01059205 157.950005484 164.067253745 177.062618276 0 225.188555496 232.452731353 191.268439788 101.684464921 166.59309204 231.30245115 156.188080347 263.056512058 230.827090267 246.053200596 218.708981142 117.124954635 199.648296493 150.772474517 175.083874672 229.136322035 213.152366947 182.274620783 165.053159903 75.0588666876 201.779058658 91.4663585263 201.07697254 102.164497038 238.290625581 155.756344994 134.105316401 145.590681016 173.078164567 222.479372855 261.092281791 76.7934772334 129.610629867 204.008380323 360.774409767 0 218.10633428 156.147787821 205.509625924 144.33152368 139.452828458 172.193084657 200.073190983 223.183932579 137.497687409 125.917700102 233.193091423 207.612679973 237.448508485 0 168.890811059 164.903117688 199.045830604 179.41771045 222.42094807 147.974105941 213.779160627 192.95980212 0 197.958454236 0 150.094083071 224.716360449 130.96876866 0 227.857062749 172.523960976 0 201.448959065 174.018642462 166.87560492 171.500515642 167.87552045 172.563927255 136.164580874 146.529605871 127.127640038 147.275625238 126.136324566 103.618685925 146.007552598 222.248946247 145.684126707 199.740142027 197.954635202 231.736516813 240.835025718 217.201071315 0 152.675506391 201.860982898 141.91114534 159.802312089 181.484962935 0 182.012880872 221.815579504 97.8853675019 151.624843399 188.39726613 163.872283741 195.796124011 112.743925624 180.547517293 206.715209553 159.597352469 166.988593386 151.369332694 240.522013245 199.562189556 175.632586019 125.86378829 129.610913738 186.045799948 254.202627301 170.417663257 204.912542347 114.376483824 230.011350847 219.018860968 194.536163342 182.996302295 153.84594424 0 202.233085368 181.390052526 155.256598478 226.914849562 172.212816237 158.067194961 189.271414973 133.322399768 189.607522123 201.735697108 0 197.239033575 223.520949532 178.512530682 152.904341981 125.547208912 160.372889025 142.501866493 133.416380227 154.345186192 228.026000141 116.292242882 90.959444722 132.161496857 148.756325249 182.389564588 217.546957514 204.477658767 204.394464275 205.12572561 200.523965796 188.044747709 171.208031657 316.9257407 207.372214354 178.563801391 223.100130259 114.724615548 177.480108148 157.333322739 250.194339329 191.032214557 191.299203461 0 169.828029923 0 215.607532439 176.049662246 184.728520203 169.32796351 217.879027425 186.331726141 150.050215627 169.273703527 146.236716893 0 0 168.208207338 215.171033489 143.933615265 162.69992121 210.107826753 269.917120504 205.018152926 177.039856564 203.484875129 204.587533602 135.983919806 128.301861403 193.368560848 155.403366585 194.061664498 206.540253564 0 155.727490877 141.005824359 310.254322232 217.778444848 0 0 115.806358157 200.949636384 174.293166112 256.506561098 190.712702744 181.500666757 145.407452807 314.565789123 157.291094446 207.624082103 136.715995839 203.425607523 177.10878211 178.256479799 235.57415835 142.330240636 233.234013894 0 219.275210327 270.460650485 160.744371314 198.514798557 159.885565261 168.295767908 129.12291908 230.428978921 195.848046864 215.690291169 131.079057683 194.621526156 213.376175692 132.311182062 148.123094748 160.164985312 165.133246627 233.956413268 218.488112377 186.317100923 209.439568588 192.100592996 193.785794314 129.200459511 89.4512536281 134.007854335 149.449585578 182.295135081 162.864315535 232.115164443 233.261951501 246.115197151 210.003867315 181.24248859 185.577021559 202.141480468 242.928414118 208.729958329 108.521305163 0 151.45119949 146.668758852 226.495023034 274.727493728 106.529628426 106.129848797 139.564499056 0 193.117710788 214.764821207 201.651895213 206.653104773 117.723436149 190.47330886 0 196.905839724 0 184.463779971 233.666423427 204.971287245 244.92528366 159.077501497 237.58847332 255.951464388 0 241.753703162 106.329463621 170.089640709 200.247173822 218.813451566 193.082965806 0 76.75208022 198.91770148 221.94714268 235.215121454 160.32056919 185.529732038 145.622370347 191.969483299 203.037163749 181.176917505 303.809501157 134.716524775 198.662365216 0 152.293611117 168.695088715 151.163362547 204.715338348 198.581223564 203.852381661 201.551955 114.658522404 160.827746004 238.992223459 169.346299208 133.204768157 153.280076047 0 81.6139341384 138.487040594 160.031827379 228.489574867 113.35352915 190.108322448 133.223563832 226.115852771 177.65811325 200.714728573 100.665060423 110.240212178 209.659979661 156.658404629 161.925501173 194.270947547 0 153.846674803 193.090963044 235.805930045 149.073048641 204.494765788 150.222542391 167.278214507 170.237368302 172.127563652 227.760023847 164.64348613 187.180106834 0 214.517439653 162.71851061 224.045384882 219.112484715 210.970015763 220.592422181 107.498891467 0 199.115089943 204.51809205 229.117265874 181.138816798 146.295671425 250.813417891 108.806203912 208.356071602 0 113.717404608 146.909066166 238.071855171 143.502616751 189.264921477 167.366295331 238.515807646 127.38422882 187.394654726 188.843177213 158.76281696 158.134672177 106.572563744 134.568746417 176.109209493 137.671897355 195.766666772 187.395249772 0 192.047678067 173.759825903 318.378909839 122.082856148 149.415122504 205.139645698 103.753893421 210.009992445 127.464464693 173.636540809 210.75006596 135.699642573 231.580637624 209.943430882 179.419892576 102.173436512 142.46264503 205.560867114 221.607214947 223.881457175 119.577321468 217.232203264 194.205098738 260.02606126 240.183079608 187.581161902 0 228.066980988 263.616466016 153.652338222 191.006293393 0 196.996575874 149.385145201 159.860561392 165.64585239 213.570473008 120.861214861 155.06864456 0 286.764898919 142.106777596 140.883003137 189.048119592 135.241766053 182.952649574 162.414656763 162.978120284 168.928876755 213.323287991 162.505455598 173.843669649 173.948648079 180.740247293 211.04407338 183.599278557 135.114583649 177.278936825 169.53414675 158.399235386 291.706955948 173.789093916 164.326865189 196.261415062 183.199833368 89.2443546081 133.05362881 98.5207060031 156.869455013 174.413480655 167.815576619 43 | conv5_1/sep 96.8349070558 44 | conv5_2/dw 222.372747311 182.17475343 280.031519397 159.467890724 0 166.723064256 147.272240492 232.449123874 153.757795377 175.063913175 237.847649221 162.083590622 234.237224911 185.908495275 175.618904159 96.3099175988 187.292747453 112.591370325 172.2154431 185.570520296 131.733661034 178.580909492 194.323304124 169.767370429 227.6821043 183.107144903 228.778649118 215.75227627 252.514057383 121.122704277 225.627625614 180.964368261 181.107383916 173.952074914 221.051516259 271.791029802 96.486277623 0 225.177744889 186.770757567 183.831824961 140.797231617 193.633749247 93.1499228368 228.427598661 174.384103985 162.832034514 175.599541768 191.751695963 287.226380745 212.609012808 216.088295567 195.421466784 177.65325493 161.807662427 144.648505216 196.541107511 152.491109192 226.902613194 200.924988814 161.308013214 176.325616368 0 137.646738781 139.474376551 203.894398357 190.741398894 159.218663815 174.626261081 152.108770986 212.943591368 125.474966261 162.077988058 184.288609124 221.363698049 170.008116012 166.175463064 116.133095586 155.671991565 183.506704646 177.288599118 193.966880513 255.469657669 183.368165708 88.7091221618 117.240193615 329.337692391 159.104985074 198.82562012 200.560463672 160.063363141 116.782910365 206.91860417 177.785832633 156.98444078 194.615897885 171.326290857 197.986233737 217.672585496 159.571599046 127.413444602 262.517488045 252.471358973 157.048175461 150.658026905 113.274263951 174.62708217 180.089704286 170.473626083 88.6326181302 254.555657091 156.557744964 170.407418588 168.02424223 208.61746749 124.792791072 266.65907352 157.376680908 247.69143252 198.789819689 199.474915589 171.564803589 166.848596982 155.544908789 0 147.821878317 157.211714415 242.423862638 275.278055717 141.305651493 135.28449625 116.5262535 206.216035345 144.721678539 0 146.932799313 171.326572849 133.384594073 168.106687645 187.272315823 163.967014297 144.784665306 223.374483774 0 146.395412279 147.635568188 187.90025768 207.612469732 140.588862868 112.946953316 155.52859493 179.355662151 158.177422399 116.281869805 152.655361 130.261522673 178.395560463 102.095792761 216.298567307 222.393195223 194.703310336 212.15214579 185.102684898 188.989370941 173.646127594 150.196665351 193.685587575 215.076296996 127.903268973 0 179.357492188 208.050781931 0 175.877714248 174.621351317 126.600608365 181.295219684 189.905383339 190.302618623 149.021335853 195.080798444 218.756820889 198.424091474 134.812925899 157.351802531 137.495636141 198.448196746 147.93391788 0 175.021702707 188.588708552 203.594064147 200.501807192 202.595853367 192.24915025 143.128628265 142.323723514 183.748841782 177.168418751 167.309116575 165.374066883 274.291571915 227.27732491 156.448932706 148.465992722 189.847060704 139.244196273 175.531447686 183.238356609 0 122.623878439 0 220.157507936 137.878452953 135.655292128 197.721923205 122.377481275 200.227138113 208.272008132 201.18094781 143.383085499 162.469640127 234.83077628 216.815930659 205.648863155 166.894321893 199.837175816 142.038715349 0 213.022850676 159.490450128 162.422787793 173.327408944 0 144.113896961 167.521450832 214.717194848 252.090391645 146.782096947 175.202363862 123.003067811 230.780148149 173.958730392 141.933143735 263.089382362 171.533595552 183.185289316 0 134.471234489 145.121116695 234.630800222 0 237.86930466 278.712832496 212.787950278 147.660986233 180.071933791 131.333190473 190.483665826 114.505380674 215.412815414 184.257867802 262.473410528 156.064470813 207.323822615 232.808120244 195.067569858 197.07556112 242.409908913 0 167.193720442 0 188.374056405 196.677429623 251.54813765 0 140.994929525 123.833102554 172.59050764 179.510875845 205.402963486 226.675353711 164.213909703 97.6122713979 0 201.834462307 175.981544569 169.267241168 205.913973675 135.236164174 0 141.274774584 208.463570573 178.058114945 124.545517407 223.356717666 174.025011714 0 112.064643092 0 150.858042836 158.381755244 0 226.389450307 160.014155754 270.534261718 183.458927187 122.047438507 195.132625958 205.548725735 126.956500374 155.495750423 225.363982446 150.649062771 134.241573688 158.746685008 153.424454152 159.703847406 203.850140373 171.813908968 165.134008546 225.601802655 214.244000894 111.103555987 272.657010209 165.949189245 175.620873825 210.781773722 0 133.0371317 158.718619944 99.8260929969 210.399704955 137.965178579 180.9620546 0 233.192930262 104.654643435 181.175541201 230.669895768 157.999339073 158.876561953 163.216797112 160.463014445 162.148502984 170.382962877 163.23147488 231.29175203 143.079313149 175.587451046 262.432866135 254.487412743 156.833831438 177.650442455 142.271155127 180.539250629 151.553957453 142.253878998 85.1279498694 141.653020278 143.754145493 174.385219626 177.914161971 131.428747236 152.007853556 121.239397633 145.177083494 210.74003812 178.463416103 199.392977234 0 156.314446546 185.547290311 214.61356831 132.252147265 138.054437983 220.393823746 154.416404477 250.085829052 239.035738659 152.901891081 126.42592043 224.757795829 210.591449118 190.508829248 162.284061558 199.383275942 141.970016181 236.901504465 243.442591778 118.055655699 0 144.399356643 138.809912405 166.848661623 140.026437641 332.018310574 171.258074101 189.284597366 206.731108581 180.767060694 144.580005617 264.285026179 199.430202417 259.446271186 170.460825654 142.837373416 189.535794772 178.872138387 155.525459062 142.510401132 145.311913832 147.994930274 136.073101085 224.772432924 172.103603482 249.933705974 174.103610576 170.134459273 129.906943731 0 131.631181995 212.950003851 114.269702141 100.581582114 189.3252046 195.381050176 183.159272101 225.527525786 173.559527211 158.797740436 212.195732474 202.411415633 133.717681417 154.766490489 0 148.245784436 190.367414707 0 0 224.470550331 202.620385891 126.205116997 154.181195187 148.987448379 146.292120266 214.826304069 162.977281831 197.414694471 172.327705779 143.319524096 166.997495325 160.847698822 120.37562818 161.746354958 118.119679807 150.435969169 103.079239196 147.469310309 163.685927458 143.507255387 0 136.647428343 201.261937701 107.020273972 145.780025565 238.939564784 204.485455945 162.698035859 141.346350645 208.380876509 214.683263669 201.700213404 158.314781039 155.169638653 227.411116561 202.526883033 125.311586217 183.577659671 232.835326077 0 255.464324308 0 189.878558335 207.024868048 192.924671381 156.671091436 0 212.852120836 228.37946724 198.032683061 192.428171433 212.27930699 198.283143359 212.24352049 144.175717952 182.030520592 111.643851304 243.648636117 228.642442505 196.033413628 144.16114613 210.889974242 186.513302653 177.462505922 136.492610221 180.381872076 151.673584003 184.028257428 250.081318324 187.432790392 168.432442821 45 | conv5_2/sep 94.4252063715 46 | conv5_3/dw 144.15654191 128.78114709 134.769167752 161.737074628 0 127.403509621 168.315852427 173.525967924 184.071560351 158.579151444 175.916179191 133.945464398 0 214.3643946 134.070078676 149.015914259 148.180628955 126.842802359 117.075193748 159.865296242 198.284702424 205.743619739 196.486440105 97.3417995086 154.868993773 212.618834251 138.440752068 0 134.505536182 143.323915329 181.533622182 146.070813347 123.217992828 131.980535906 0 145.979015723 170.246644147 152.055444773 153.671272164 209.288945558 104.869740782 0 118.609193507 169.589995373 164.637295686 223.29166782 111.36661381 118.829454118 185.052203938 164.343819472 218.429016749 157.538036435 201.657180796 194.40130616 200.840978345 174.59434514 176.382797286 188.77492148 207.754631629 165.560440079 172.889082674 198.333452819 231.755881293 169.002370795 0 0 183.301931422 156.974934937 136.109094854 127.251547853 0 199.541223934 169.76856808 182.881992587 121.476701522 180.132841815 135.31442712 144.204934405 0 0 157.837589185 142.667445112 130.856739915 98.0540090082 143.635777265 190.073532757 113.742889458 108.057471069 179.54438775 156.809715167 139.013840846 167.625832417 96.4217288523 222.172921137 162.828472289 176.516143901 0 163.501099511 186.845141644 128.623187786 168.976867331 0 155.995597726 137.97809123 158.65730137 138.624779768 104.742862585 191.557087444 0 295.885596148 189.832201946 176.063015747 136.609600674 105.700828104 111.79027522 0 173.827304966 138.95350217 0 110.196126372 193.513042981 166.756350387 200.077996484 254.077832412 158.517236757 163.927614953 168.484212328 0 222.546430909 191.201083819 142.500915242 203.126526774 0 142.434519934 179.536061583 179.585029903 205.341321148 129.064669085 158.018159829 125.904017897 174.719135192 156.662148813 87.2547819692 304.8626228 147.290359671 169.8431525 111.751671179 221.470607237 104.981461387 140.704605609 0 177.919705525 229.638027874 115.805579522 145.699228438 147.458521995 172.801933762 122.225645553 126.139372778 201.285257594 181.218392903 173.69510301 212.913362851 112.797726152 156.548393642 174.849882422 129.26016227 182.179296443 0 114.329965672 168.510703438 209.823043155 299.90786279 171.079750171 159.078833479 137.264713471 181.201461923 121.123874677 147.252182339 145.609748632 143.292258998 157.114780985 186.055908084 164.026920594 150.436063421 198.906474456 171.254274723 155.732071119 106.580899019 173.488349039 149.414589312 153.837724553 174.991203401 130.89506211 159.392839207 147.499328878 239.077589955 174.481077862 159.717070425 115.201663718 169.679209237 146.130707634 0 193.927915379 185.307308911 125.71211285 154.840773184 213.001111509 166.250170137 121.990484094 122.527772115 132.746664133 171.229038527 124.913478261 156.158153252 159.117955217 231.051490598 184.489156576 195.859927983 114.934241812 172.616895766 184.520839219 193.382429252 152.30325127 194.532182325 160.677097278 164.521506984 166.942285088 169.468517528 88.5597791359 130.207991027 202.672205169 152.056402421 0 175.551575076 165.87873365 206.858216027 150.408902944 156.057090835 192.07752086 121.691034369 226.169516075 185.838996432 160.842732094 129.451994614 159.692508317 147.068629673 178.130292317 224.156225713 125.847032437 172.936315749 158.202604285 156.237418096 143.298650369 145.130767979 149.274345997 168.655174897 225.613206927 143.477877422 248.371131977 204.394541153 319.186557745 161.157452213 95.3844973412 187.06503126 157.815719665 190.128319742 165.229676959 180.650078982 139.750541639 156.852563377 156.281251983 191.32063019 152.970030099 210.34396528 161.34986597 118.264271042 176.575396412 158.068232702 212.020474752 159.934320417 202.553132638 138.592426635 198.617133248 215.67778267 150.012208919 243.855251256 0 145.453272649 0 136.972921601 172.554021611 145.112560578 194.957736477 115.812887877 201.128016528 156.541099382 0 100.967540353 212.11489151 187.13700548 129.163043652 160.382200518 121.852491315 142.973348276 0 189.988729 175.794031903 179.97609402 174.744469353 214.383536993 152.179595038 174.131156573 171.926323634 143.463579702 108.889191155 198.028745701 200.757568309 145.365789872 179.093907603 145.251364265 160.984910844 134.894411476 198.982028978 251.040298672 142.309094988 182.912752861 0 175.559000456 158.207503387 242.842301478 128.383005019 190.74077242 129.410393554 211.541915483 183.24675743 169.196011298 0 189.510302722 171.759108929 165.909645319 160.098426935 145.425559567 0 175.792070287 119.073942555 248.65805209 141.462819614 184.749934684 156.803746592 187.603884008 185.371004056 153.411924374 205.787937537 195.200073746 0 190.411416831 183.558245826 203.07546589 172.506174588 185.131148732 192.482890962 137.978749499 179.457374234 128.536871921 181.37461863 183.068287259 135.647923972 135.724356609 205.990045673 179.109881034 125.007193324 203.281838815 103.803642415 206.186988751 142.819704939 0 198.844840321 154.031540254 153.182786777 166.621985359 278.420303003 117.974681135 231.250703833 127.335391564 165.517368733 156.941262104 205.205009479 187.117483662 282.933058785 172.011771303 125.760779721 197.427419764 162.070631784 151.232921498 151.433749543 188.728438552 170.645032477 182.998739754 0 131.752010997 118.329663992 90.0241599785 155.642265199 154.902943047 134.137413925 201.391226957 190.224036034 152.568719395 162.862380098 95.7668458666 0 194.985081556 227.08629934 136.503974208 206.843371227 219.595821313 217.775409424 182.849010696 235.718351276 172.703191752 160.289282385 131.263287583 163.68453137 183.962464698 0 140.492266076 113.039164614 174.066576517 141.347176548 85.1946896737 188.015706121 219.830392204 258.188318269 175.739266151 191.762188042 169.126501063 191.902379512 131.988529076 218.79296111 168.323220906 144.441479647 152.640340509 158.604942555 153.295143823 194.439129 164.159737956 192.906112681 175.865029134 158.581820373 175.111259594 170.924184899 118.160335812 160.611789841 159.655768052 193.966763087 201.672466993 246.714890872 120.235884225 126.716838414 118.727177557 177.780154765 167.566019426 162.787151153 137.335288795 151.656649974 214.614100877 194.983902858 145.374702329 155.176252539 167.030641254 97.620195247 180.900287881 142.442451132 155.27791917 158.463072555 132.713347611 213.653585203 176.798966946 104.95156597 158.533137119 189.663318493 185.968072182 133.102769382 116.836066352 151.825195559 108.001362977 180.685645565 84.4909484213 211.241478002 0 153.988222727 137.831889522 230.648027232 209.753606952 129.183400041 221.651990463 205.254807257 209.247345878 182.620189773 213.42051328 153.350716762 215.519970033 103.618355277 172.883980721 133.662650688 183.346019013 141.770231182 229.578838039 153.926577724 161.140846421 193.751325097 47 | conv5_3/sep 91.9809348677 48 | conv5_4/dw 118.335162047 146.308319326 146.830822144 160.702825881 108.506502266 124.453285791 148.214171533 210.702919554 0 137.616615007 117.438248465 158.598976116 146.842684651 115.089420148 138.988747581 94.8353198587 140.272333173 163.273925982 0 159.179205918 109.807019099 92.7002308513 0 151.919293463 147.758362366 140.846054518 95.3261602254 135.747469951 0 141.048774391 0 0 228.381948783 147.1971217 126.050037682 210.519546088 0 157.49364093 139.408317357 129.6242916 132.713041867 150.620612452 112.488485757 216.646158053 117.56167078 153.20865629 122.805517294 135.255898959 132.300917186 135.101493696 123.116865706 148.536492128 114.447388374 168.62091266 110.894734785 145.076216574 80.3531858022 123.322781333 247.445978192 154.253550254 132.134080259 139.180698636 126.253447108 151.274308328 158.919971088 129.160091475 154.593328075 205.900896792 149.795284878 168.893949795 140.099896865 132.728163955 95.5121645639 165.569760664 137.898180041 0 137.882630537 143.200270379 172.118777054 173.851897894 135.877539071 169.296588576 175.281717009 214.767268531 94.7749163732 107.757782345 0 129.652944789 198.256086605 166.279100226 140.31470185 65.2540702556 152.369816398 138.741170868 127.999940317 148.269329655 107.033545542 150.890004794 99.4767184545 239.95179816 0 121.736825106 226.58673643 120.127954862 136.994194063 174.692738945 205.760636274 168.383729114 87.6676932788 125.765102351 143.575246324 171.113409005 110.892447659 163.960429585 161.295418293 113.241612727 154.666459779 138.483776052 123.606639987 197.39922561 140.877079482 0 184.511588699 149.770209676 168.447916436 109.775573391 0 180.023684791 135.531459951 130.684088929 116.974795993 98.8825039252 117.487301315 107.034320827 106.949761514 129.449929589 79.1875931293 171.843900564 138.987975288 137.275833234 151.191771298 195.909062796 88.2176253448 119.989580052 0 193.713608089 134.940534929 161.168548746 155.827126409 137.10397387 0 98.0501670463 117.293520347 0 35.0691154349 127.913724815 174.67176242 292.073174255 146.525584473 155.068815999 114.663294761 124.586232683 0 102.330727129 109.144191129 173.748363838 144.223454426 107.277900097 116.976227587 142.100213526 0 185.032391227 83.7781102502 160.323430074 148.936903587 116.098495985 99.6831137955 176.664463499 132.751203758 112.347714876 154.630091449 135.514594243 152.227089993 135.724020614 166.55543076 93.6871887327 123.028513624 110.113864341 84.6331778284 173.768773935 148.768258703 147.402176231 129.903072675 150.480411006 132.148432411 168.020134417 147.586638112 98.0693256598 191.450632449 141.715869069 120.119392656 163.819582115 141.636336191 0 127.484481316 138.085694724 134.563651987 76.6060106109 312.156815265 78.2948421834 144.489192844 150.318926568 133.04159352 130.128459569 151.627878038 167.13179827 168.603507806 0 192.869552696 0 201.377660475 141.692019997 163.526463641 199.484956343 196.573927855 157.511800467 181.01195155 146.93019286 148.367207117 177.501791749 135.215281601 0 145.002323534 128.90830647 154.525264878 129.288211419 223.115098814 189.812841462 143.803956561 94.5338613819 150.416760523 159.721694098 121.004093714 150.185460835 127.562390401 0 141.516546961 152.026794236 182.316381194 135.87247316 179.033464895 135.658680805 140.751840233 140.457888765 221.021781316 161.705911578 114.654281431 0 156.683680933 210.701752273 129.45463749 170.811578353 135.546593188 143.387671313 144.911861871 140.745699299 183.653562088 154.969819803 121.577110258 192.299130944 162.402939337 161.076442684 192.200775842 137.092131145 129.829300418 152.16810047 125.49782173 169.029325159 171.831847708 136.610816747 104.76405591 174.467695803 153.426771279 211.041817609 204.640185631 156.621843526 135.32133641 106.929535039 133.423703475 0 115.967886172 119.5052587 191.715907931 128.429152437 148.812216353 132.667182536 146.986255792 185.147584678 94.3858400785 161.583373445 154.574871258 110.343955809 163.373268491 0 213.10691612 192.623259905 140.560797767 73.6153166685 139.935086594 131.103864595 153.85033813 0 0 0 186.662107656 179.156479771 119.26312281 116.171800644 176.804883304 140.136773618 156.997885682 163.694764844 132.427081086 150.177108702 144.441184186 116.194871126 112.443283628 136.192935883 142.005705588 156.332614274 154.408694555 187.756251278 0 121.282648653 228.312091401 78.6495700947 135.148976286 138.8072778 217.822422771 146.458635939 159.951755124 136.741681286 145.197198703 155.364715212 137.996320938 0 120.235841107 149.349236383 113.396836907 142.182242639 119.594243008 153.977908825 194.792390629 126.915334173 160.217533043 167.34374774 155.142550856 209.307196964 168.957526536 188.259009168 187.479397493 210.36318624 106.664479204 144.481422125 189.805632504 123.444162445 167.425592403 123.711403597 192.406440526 98.7469517783 191.531081988 0 133.373331164 153.325036861 188.474854173 156.544734882 111.995354125 184.63023862 119.003362281 169.287332245 130.544605841 159.813960426 226.879185241 185.104472176 159.169839157 144.818715587 0 160.937992366 157.72768863 162.068142257 0 126.962330285 154.801840659 0 153.080372402 211.896263539 134.369572395 159.757973218 134.819032188 147.452949416 142.341613452 0 164.380480001 131.403859744 142.35211715 184.51544605 113.241037457 185.322515894 148.005569559 117.651704097 120.384799262 132.415514743 130.64530683 137.515130325 228.193182499 143.120186217 125.832027631 210.976655938 170.757961123 132.824726774 106.157917106 132.342592373 192.86637942 175.144223305 135.516344131 109.04406669 190.291945117 120.888891272 163.549518525 124.436806611 201.189518593 163.936849333 118.819909 82.3659851905 306.174924746 126.132407538 243.503480198 180.655757738 106.78410292 172.900090579 121.750244305 0 99.3264432976 173.082075793 100.09364837 182.599783749 129.120486204 0 226.934457401 150.666687555 154.466982973 189.398331408 138.356234137 82.0131996419 147.959701174 113.959182524 137.913938894 167.682011414 89.9710696431 131.696148536 137.957534584 207.026204221 127.619114253 156.753036263 144.324658692 169.935155494 170.076827162 0 103.255920258 0 192.981900359 159.927887963 43.2280647622 155.807452225 174.90676258 116.370283507 215.537715087 219.712542525 128.361198611 132.636736023 217.249663687 106.415603991 110.885961386 136.564445443 181.685611984 168.560988873 134.180931282 148.325185947 155.354369967 164.574249548 148.001511688 185.02592564 133.680596117 0 186.135836883 76.4011503139 139.45906271 121.840165918 148.377606662 129.406362782 203.232649855 167.506045821 133.966069709 178.724612037 143.100912034 124.830103967 162.470972719 101.354236158 0 177.386320082 201.614111769 165.995240483 49 | conv5_4/sep 62.6347921008 50 | conv5_5/dw 116.861091468 107.838747119 75.7585309079 110.906429139 197.223495478 93.6779993114 133.053080629 122.26003587 104.696427284 0 79.7073679142 42.2581345161 0 126.811626298 162.328150625 175.541166021 142.887592765 55.3064308714 130.109242435 209.377373378 137.900936476 117.241698221 47.6428447443 158.793349155 80.8311345672 107.294171674 133.033866403 168.729746508 121.580576467 131.202791602 82.7028267877 245.011132467 105.653760871 93.6738843344 195.879527093 54.5858864865 185.809879761 79.9874212184 59.5965240337 141.831312422 200.513750813 121.524221845 76.3999990167 131.06889007 156.391021484 120.721843929 52.2796437168 217.45733426 49.4358172446 158.58184027 143.559183235 106.827263368 102.258568073 156.885240997 127.981858058 128.457141406 88.1565276728 133.488048838 148.304060321 0 70.519222095 82.1344783529 100.302705349 66.6359255226 115.537321991 198.635831051 194.998198238 205.10900716 146.309847887 119.93890691 172.191308938 119.998897131 42.7900681294 93.5594109126 103.809182752 85.4660460547 56.661756393 120.959489884 169.756926294 162.572781817 152.195822192 128.065883968 113.537781531 34.7480231068 168.380974873 150.22713289 97.3624474087 144.119116351 88.9839252286 144.328368081 130.128859766 140.374826531 145.960693137 139.194306164 92.7292703191 70.6799474876 149.657066772 81.0315131507 269.343601769 112.403270298 96.9133603999 83.5044322078 134.534483105 132.385038042 128.145225685 126.500130974 100.979319861 121.429721846 126.01968045 118.380350093 112.408736493 134.408599212 195.89768639 123.363618257 135.332321726 143.76890118 126.653996929 137.697365834 135.641960314 172.409853956 135.597837573 0 154.771306113 121.892165124 213.919419407 131.620854861 145.896697395 108.224454991 107.770468289 0 61.355937167 107.974878457 144.662218349 70.7200512006 133.0011905 143.563430746 83.3072758857 177.641858418 164.1026976 145.667412547 137.744605602 142.49979491 125.251718621 99.780715109 148.807186124 103.839596429 91.5151924034 100.013563561 130.709110043 118.814266091 68.4969776938 157.135264259 83.481102195 145.332284424 116.876323404 140.811828818 112.762195597 134.406725216 140.652545978 131.47057849 116.495317316 23.1714285752 73.5263530791 75.9996499835 56.1076846297 169.024995693 165.250830016 110.106280381 101.967309296 124.082231182 0 98.4057953768 0 113.782504087 128.479951788 68.973114591 115.602530654 132.191134075 79.8826563314 67.1564644801 66.7037835758 109.576371594 135.992187441 158.711275943 135.784671655 148.344003772 117.365799465 49.4920923321 128.215496973 88.4626179328 177.637855815 96.1783374482 70.8530026428 67.738376603 112.089617593 114.498105914 38.6589140674 92.7973472693 59.4353286789 104.573472012 133.641878173 146.587015404 69.7228418834 123.006569598 0 80.5365638616 110.946642901 62.888671505 97.5979959589 0 134.083502496 87.9615355898 126.150845907 109.320422147 188.441145277 127.126286213 67.5032778854 106.000233191 174.939835132 0 47.241107337 144.756488958 142.525882087 94.9296079386 59.5458197763 49.8829559865 29.3799414584 145.235276729 191.426553397 128.57523312 97.5740183548 43.1984531878 109.330981747 0 88.2701916688 119.555817498 127.059588191 108.325466395 104.294647983 118.515839512 185.968918484 95.6446603842 59.0053924401 104.587180955 111.856092377 146.981900301 113.848469706 37.0109078736 184.730630807 113.38473261 180.215352245 58.2751890395 147.181038645 118.126574495 106.868496683 39.0218816023 107.99329356 104.071463194 62.5444974821 85.0385832726 103.983918644 103.747468733 124.561768194 169.420858896 96.4621711147 81.0071594421 58.8850556752 130.111033746 156.528196835 177.587332068 115.267688088 89.4083556214 94.9907267402 163.684109977 123.328447559 150.920306259 111.093025462 79.6355751889 106.49201561 167.979129151 194.444401425 119.838330929 84.0615381157 0 103.042690547 156.470375483 169.571937128 87.2173355675 0 122.066396078 175.372050763 135.599002792 129.415137763 112.451414208 109.318253829 140.869437516 165.982086029 137.829600283 61.7504457992 197.809631416 136.82576319 86.5298953398 133.91762608 162.084753694 308.855952048 141.855156668 116.812475556 57.1295772389 84.1105556966 88.3237171595 15.8089387216 109.258547568 74.0278641173 0 93.9139492167 146.252252608 124.180605205 134.919735118 118.769782702 66.5497553668 105.13763101 162.597747027 100.436656824 95.0787562837 88.1565880601 83.1826993674 59.0920861771 157.484542924 134.353887309 43.831327875 123.034833938 46.2021811094 79.891878082 135.026574549 81.6524121668 0 104.140488175 87.1778350832 67.2457121889 143.145578115 184.550333458 211.005604964 155.727097795 96.0715394204 194.717350271 99.1729192986 78.099264408 106.213180821 85.8124449577 91.6789534893 65.501876045 112.362926602 197.223116869 46.8981931811 90.9354609998 0 70.5083385958 76.8621649047 74.463050062 73.1921236299 0 61.1666611268 96.7396789381 176.656958417 67.8902447538 80.2423672487 115.367607153 130.306061117 90.5594570328 125.07069428 0 110.606119256 122.779574138 217.711305049 107.899502958 0 110.058647238 168.16862889 189.595791221 145.270191289 127.056021401 114.088425824 117.185306325 235.699703428 138.425562847 80.4475410364 106.411973159 98.6696153824 110.555130785 168.203225867 114.583948104 156.771843552 188.14339877 98.0123881186 30.436112267 65.4425571389 174.960258987 130.76236396 0 59.7515663642 0 63.5686841107 121.392368963 122.263075507 101.607523083 35.6950968835 79.1773195214 91.7223618073 105.897233034 114.629864479 113.259604811 125.018449935 160.951421393 60.9282095257 166.007502476 173.15434443 106.302810907 129.582940138 152.272638441 106.449316257 144.042105709 121.132722779 110.60673363 32.4666125348 85.4138064847 147.28340534 98.6829292292 112.811443169 111.583920155 82.2222716962 210.119823813 154.427630114 120.398731727 186.598085139 96.7372835624 150.068695764 333.195595633 51.5088073131 97.5433167396 125.082890088 80.9803236106 81.6994115568 146.570128076 73.457381865 164.609792283 137.201376746 115.45491486 63.0533923866 173.987261214 90.6321733898 134.878422769 152.550046281 0 95.8992555091 143.132253908 121.247904279 186.505245891 138.354148409 228.640207845 118.730934613 104.557058372 102.444056348 172.605075503 0 0 81.965397234 126.71468405 116.158567915 95.5986053637 100.092791897 100.41355109 0 131.008737427 118.170440881 91.3265417163 159.603271333 140.066848089 0 108.550924512 63.4498913871 111.318978096 76.1027118142 91.2359526745 110.61650586 127.60737016 93.3776041056 98.29717715 90.4185617355 98.1914890672 83.698497892 0 135.7043014 87.553226143 177.893322631 105.347775895 0 86.0148942975 120.814199747 159.298389707 129.759643676 119.502097696 91.1419240745 136.364276892 171.967855384 0 128.470317533 129.656143044 145.197506017 70.78465322 38.2171570128 163.378563918 151.097650537 51 | conv5_5/sep 64.7028861592 52 | conv5_6/dw 159.924762902 291.577047663 31.2551735644 98.6603143773 63.4948983235 146.815194463 220.073822493 38.7484270142 43.2736543084 80.0790053258 145.726725196 110.889994687 213.347708055 177.143795375 262.641724957 0 125.274540352 96.0454593924 133.649277088 212.947789087 130.338217814 127.260276314 0 28.6677415435 26.5374365437 258.179487848 170.722292751 54.3864436973 169.254257927 145.757822334 123.192059653 167.6939739 44.3399436456 251.464848163 186.48529151 101.175250199 93.5218244429 0 177.596837497 33.9902253562 184.095431944 60.9056237909 0 136.86753223 39.3136291111 30.2578530087 161.548472803 141.284337775 0 0 36.0994394769 26.2476670777 117.949459655 127.736603772 154.307941794 163.694775639 127.957234296 117.840026989 139.678912348 176.82704433 169.194478346 0 130.349559096 118.680449577 240.202346902 84.2995285751 203.536949522 126.895437537 184.589823434 106.055338957 153.254028685 0 118.570435633 86.5568122186 88.9628688371 207.419554922 29.1904265841 128.607806478 24.5108096468 106.922972025 179.732031774 147.365992376 144.862100002 199.929391491 212.223467945 167.339350493 130.803689577 0 51.2591022485 115.190559998 167.614770996 111.003885802 212.991651596 133.232170674 156.947739979 0 108.069552987 157.005299963 139.911689507 106.523127001 292.405113801 187.951167543 248.958949852 0 164.852684544 215.381862357 42.0476250836 44.2223768489 161.39444589 110.203294657 169.488197371 111.319539645 30.2136828758 159.345772217 150.998057793 156.075402116 121.260430604 137.224869247 31.9042450737 77.8948052889 162.510508934 62.2051479805 136.454496362 139.79884617 185.558407342 84.9973542311 30.5371347068 121.619697166 173.756504525 126.091798852 244.044990198 87.5017827532 205.908024953 295.782925816 105.802823776 118.219913719 126.05285791 173.549360302 42.9685146427 132.610699691 246.485089836 126.676902651 0 159.276217861 47.6947915996 65.6835177649 158.156137452 151.70257792 151.178011513 214.850676465 123.145297231 150.912029408 102.471214237 90.6276633101 86.3694437289 184.67036993 187.636766661 117.159963594 189.65280254 0 35.5599821536 121.68769781 133.647060836 138.343771031 189.940834689 101.607007554 78.71926663 75.6627968846 129.777540926 0 109.699290134 298.577772029 142.21823448 37.9448301719 142.356951024 154.737275678 129.036063174 186.458984767 207.462394243 203.103655744 149.76462239 0 137.263251096 151.21854323 0 168.994416919 171.864365118 108.137053972 92.760309811 68.9094479821 188.880015965 79.2080921261 170.752572905 73.4409689624 190.592460779 107.355802817 113.761464981 202.504328565 121.593809895 101.259331782 157.99241631 182.442596402 114.641129029 128.444163873 0 44.1409200335 113.457159347 141.267042042 29.1670309143 270.847114202 99.8587904442 100.6474198 242.776748303 108.525075579 130.948598569 119.817400754 197.231350914 137.26347451 41.3174832035 71.6399384682 132.16337082 162.996958816 184.945310883 211.346215866 116.0754396 90.0886995715 164.830319279 136.723746627 168.953177391 199.481011601 176.418416346 226.649281932 280.174411266 132.485858972 143.224386581 339.036085468 109.431029107 184.585971323 156.686678914 127.506301685 223.314980637 105.515512786 34.2839978091 174.578786098 96.4232039488 206.277677103 125.920499812 192.669932714 115.407103744 0 181.517380903 22.1376780563 33.9514936822 88.685070704 29.2075020103 178.626640403 127.738428222 152.559291333 137.1713775 194.727876537 174.699789643 112.679499762 176.673828397 167.429773808 120.712458495 153.02135722 38.9740824501 181.053137227 17.3921338374 105.216433858 152.794172418 191.654685259 31.6782711208 140.800717262 71.8732170059 215.195072769 38.1450187678 194.77485167 100.404014631 105.539264054 47.0559849116 101.969987341 129.136054108 177.619207849 123.275244152 0 27.7923984175 175.914247857 123.189801517 196.107665076 196.556254137 92.7247218803 45.7712831525 147.321970862 123.914446271 196.100101745 114.255717712 258.04423941 81.3277763708 99.0223106181 17.0985829536 217.916418202 143.992830531 59.1356591912 128.817524892 219.279172574 0 116.526679599 153.628579287 102.177042679 115.408147966 154.664002785 78.0610627239 0 109.312088187 129.782926566 118.790807797 178.23123708 55.6975239027 82.4019181959 44.4723786888 113.067568132 0 38.8245268044 94.841006237 233.019520596 121.771590384 235.770985633 103.781377519 109.341426048 254.6751439 119.38735722 138.814351919 96.3443535437 133.151540921 162.99535871 0 155.80151381 39.6920090806 198.004054938 245.462221095 75.37210228 160.733850212 39.2084345489 114.042261297 139.947786981 168.578592104 183.566324655 79.4869219341 226.110870739 79.2638150762 40.0713225033 181.116146202 132.161631727 168.685017078 115.991537303 133.313698859 137.796426089 0 0 140.854902402 192.569487241 139.389592828 29.3113202539 54.1726826734 160.410785125 188.690409935 147.275181719 72.1261090282 73.9260750373 138.000471741 145.792317838 107.988752467 110.086089431 184.218288736 95.7289729533 154.553963353 314.417107531 159.220013528 119.039977383 154.113994505 0 92.9468961904 192.275982387 182.282180488 214.424617063 205.443244028 203.288188901 53.8083243006 339.870361794 95.1942761731 137.932111737 72.0944896573 100.713797505 132.509882659 128.327383676 136.618989554 55.6778572168 255.212080334 228.031000921 76.4801955835 129.62088606 194.079923859 158.486338699 119.828414467 191.294783724 210.768683979 0 71.351416252 187.738533715 165.913697536 175.238943678 229.172136541 37.9182161103 210.565184022 167.434602174 0 199.41849387 34.162746963 123.956644821 109.882713839 128.486842693 153.622706638 39.5833567503 222.93575716 0 98.2602787925 120.269264737 61.3265091283 166.448939421 119.526898209 136.149481259 157.267542138 103.454635435 137.072524796 68.3285479325 272.272867559 144.847790057 121.057174409 161.586463354 100.135332329 34.2260539271 97.7936842316 41.4511917324 169.437387749 0 136.687818097 0 129.506655537 114.407736984 0 195.673965507 183.075795694 154.209046437 127.221556335 0 167.800552603 53.7091383374 114.823608638 111.262301909 230.595691948 54.5856526211 152.633866914 142.591249878 122.137824063 113.802048986 87.7721635782 165.325030447 101.1930339 245.5543187 122.500130515 123.562494742 162.181824796 265.490679916 228.888767207 187.452803578 38.2740987827 0 177.244483787 137.319353738 247.589456711 173.197248074 156.521348252 226.445178769 20.1261549643 186.539692483 226.679854091 143.243032129 147.810002548 52.0899486453 169.093328251 276.454581013 162.770662769 67.587873329 110.098293829 81.3257292798 130.82526806 229.675969076 176.38484919 131.512563747 212.710640381 174.569893138 33.6742527246 98.4693297761 137.036605319 0 144.268265223 73.2500671725 247.53097039 131.530707212 131.247165402 125.483016152 53 | conv5_6/sep 105.72693777 54 | conv6/dw 150.654990839 415.432270709 179.926273673 361.600396089 335.652521196 434.059736223 184.112824013 250.817030035 374.870871496 0 300.055060852 359.670419217 0 340.560472671 194.14311423 403.412030835 256.557495907 399.831911696 65.4247523829 301.178864412 0 243.69459032 272.952259457 320.111108096 461.245019375 303.318280348 163.596084998 149.518395436 261.554238907 370.670334627 163.45902475 337.319940908 159.732447483 332.079933023 334.153609807 305.507855604 147.498651807 324.15106104 296.129710583 278.994629983 319.234292614 432.750607678 362.296952724 230.702227137 272.418153423 0 336.055156585 230.157087636 281.44780256 390.59371204 419.824318563 0 299.353532664 300.987053246 339.264330835 412.278401759 364.103988897 282.610367142 413.257008953 437.231450792 288.34906471 0 340.887372638 500.778478322 144.665382201 235.488978945 220.832789869 144.313823723 188.717589363 0 206.788561584 0 365.890045072 137.109951892 0 385.802681287 260.889492687 251.306952614 356.792231734 356.953422426 361.202927816 332.697524487 369.832893554 141.552816499 401.463566907 338.122622549 325.015491627 322.244633582 406.280668991 0 271.649735063 422.860193177 334.64648927 334.58469034 141.306954815 304.345682153 411.164016033 383.463168993 179.845907415 371.416094128 156.55533634 333.282482205 311.1624852 0 297.057689827 212.919103149 0 351.718115613 385.514236432 320.401389229 387.99484546 203.17234704 478.354783393 369.894463832 413.565008662 414.080680553 487.988931322 408.153951236 194.578212376 131.364931343 0 237.023796356 390.637953697 0 371.671872332 252.274092735 409.671798597 356.383307372 532.773995259 373.905685883 203.397206454 345.834286581 400.976372509 344.575773556 274.83496888 0 375.021954495 234.009073736 328.289125252 0 0 391.612733447 368.611936055 376.231371086 305.034518124 228.795913538 336.827839542 275.331394553 200.011349801 337.959779789 275.106978127 155.70636586 186.002757037 159.297422662 378.413891273 377.672835727 0 348.663051433 329.770544269 0 325.162437057 0 342.753949968 309.054930756 158.744396112 329.355824102 0 282.513484077 120.020328695 259.828299532 218.157290296 352.193077189 390.063651035 0 305.805416868 337.194138348 235.320469108 335.396644005 261.082628116 381.547376112 402.943523563 313.635858089 375.784054269 375.083788495 245.873575432 272.198745355 345.028145886 441.359791887 0 317.490105148 267.077933947 370.41578712 430.705911924 287.787231998 209.680680193 303.570280662 416.095217451 148.415338191 487.673155447 248.189849392 386.026101275 290.468557669 134.958760823 313.532819504 345.270777052 366.743508721 333.217752511 215.500561396 333.438981634 235.808964196 295.786603068 507.042612631 330.599785903 319.714850587 237.629481504 285.347711643 309.239818786 303.901660261 341.72911339 176.117329457 359.517825259 317.814687502 327.317655229 317.694684057 293.672835831 348.742605386 447.23766152 88.9390284085 358.359931452 0 289.784598614 241.684132492 422.868509334 287.148853564 0 367.421276513 376.220784633 314.66877064 311.201477338 351.145425755 357.522945917 358.615392762 421.136192437 458.20798665 401.66241822 404.209298251 375.156112139 0 315.292228348 208.446191351 308.341889853 460.369686824 275.698227905 249.780251505 347.468624851 354.407232418 291.863310739 158.639986063 156.330851213 500.224630133 165.569541118 470.284922845 304.230648062 352.520944317 364.827112217 405.684602994 335.674918936 317.777230819 315.060698338 172.283755383 354.44505263 157.557787259 350.596981242 334.836034717 339.123718526 172.230647355 340.522715134 278.217215407 324.086491648 170.333863589 317.524754683 288.484968668 0 224.725377994 321.728191341 351.123576569 335.491954653 350.723353576 170.882397657 334.708219521 280.693322123 69.8257826437 342.169186322 114.365982901 0 298.332369586 326.542197515 262.577692871 354.449629492 222.779850038 427.201457886 233.470235936 185.285293296 417.491349205 0 0 276.998718783 187.141389132 240.785405969 266.640263642 200.281576562 203.960713044 341.896002926 218.4578629 353.975844474 350.01688782 493.689230068 282.997460833 198.266356074 292.34726528 386.079800216 189.497905718 144.343464309 396.276700224 152.182291789 467.148262081 389.721906823 229.244112781 347.284478238 316.531822747 261.082205805 232.111197604 203.52035374 422.154191219 273.041439582 207.239743894 343.151812332 163.010952874 196.587834182 434.050714766 400.941430794 0 356.194368357 384.323880204 333.399834546 359.065730816 388.127744325 245.84287521 285.78494363 427.7681949 375.572366407 308.822528857 316.654126609 366.982001391 297.409582187 354.454042784 449.265991022 317.400187278 69.3681969615 412.500831584 111.897240002 461.567838573 212.022630586 275.245651128 275.528341754 311.793848465 265.391783048 373.815556086 446.076178232 342.523195292 346.316630654 334.726378446 360.235613882 298.556203915 292.139604569 398.708186925 268.518063494 71.660041113 254.188601085 286.045342576 0 206.573308946 287.425937235 394.210414583 134.875087832 410.447389656 245.706275742 0 374.848700394 214.570126538 237.527527505 192.499797153 419.838819312 519.67425869 402.951426512 0 159.995769236 246.794512898 137.276880716 360.629078152 237.657302595 458.634189746 271.879748344 188.357009769 400.567930502 331.749756336 268.137102958 0 142.230331643 140.108766708 297.234709052 0 276.350709402 319.521707647 0 333.364940301 282.608814968 344.567406614 339.716295134 434.323072985 384.264904792 289.202701609 303.633918839 0 423.803890395 291.150264143 302.893068615 58.2523560913 188.19885114 338.955191824 0 0 410.709570905 304.684180911 273.731845204 319.020302309 298.226208822 308.935502757 355.777136346 215.49478683 299.514941067 349.867991436 426.986449921 200.293868226 333.127351418 334.653063004 193.35196089 277.178551164 533.975025554 169.441640397 396.688997464 357.765779533 158.883943993 320.288294603 279.371113166 241.477014067 336.560721065 51.8449385713 273.956521745 341.192023444 146.194784483 402.504011893 330.229583873 235.036049794 340.58513878 393.05092578 308.734649949 401.636620515 406.018492572 582.757745208 377.491019767 144.907368819 442.967171116 342.48180642 155.966049437 281.261820358 327.34764584 361.571146445 174.097835967 415.218066585 253.981975687 333.612278044 426.440570628 427.777359218 355.30487443 0 275.800439748 284.433807149 360.407338649 267.813749059 292.013120623 268.265450651 0 330.096698644 266.628896747 302.357814632 300.491445031 346.394880112 451.478096777 0 493.612631614 259.394299216 432.634473405 0 0 184.281578339 0 242.872876553 257.229500954 355.7332939 149.651479398 355.399004607 328.673371192 348.991293478 344.505888058 386.084246954 368.018951762 368.719668425 330.450888319 320.808483046 349.065297253 272.012873999 275.623552852 300.393533097 339.677626237 147.22118668 147.249599545 158.082435384 379.196814129 299.293755006 417.111041169 380.805163193 457.261946502 137.964788242 178.965533864 286.646141896 212.87566867 314.097467487 455.07340049 381.203548765 254.226623355 0 440.194232784 279.852297064 426.173050932 341.974818063 392.637512548 197.905800977 0 321.515616737 306.717642758 188.185271616 385.560775651 405.912009292 385.965315766 0 381.658675223 210.354201551 0 175.349309181 193.344477251 330.752801374 361.343390248 140.211885002 149.999047952 264.023282753 155.282802052 251.873888092 250.689214569 0 288.502409946 286.650933061 376.583862 347.50641232 253.742017861 325.708451832 361.890598781 191.924380114 389.614777591 410.943677098 177.829074936 333.67232679 244.060716609 220.59899881 355.567941958 321.733360861 0 351.538897974 392.77591223 335.304472867 412.03574747 245.11029116 158.974776543 253.816524816 184.484574711 247.337712868 254.955059524 344.0848017 344.233528766 260.445606701 380.273262642 388.364484631 381.300022351 151.750791611 44.533029583 266.14357017 214.796820278 190.336525567 357.374422581 0 353.257930338 399.849629779 310.990711718 302.047654118 379.092050419 234.506716112 419.429234558 312.876067807 269.911789031 331.621570135 331.981155313 378.642676998 323.665482396 320.850585046 364.24488213 578.769171578 319.367768759 183.179079548 245.842700993 433.32273799 0 0 338.875958545 192.757194776 0 379.917697023 327.887091754 160.498023361 304.416265012 0 451.637723589 184.720203674 306.527716704 191.464060599 388.634099465 452.901915762 311.06205553 424.76736033 415.49422807 362.015647973 170.283886343 0 353.018508996 0 258.257858595 262.528810703 291.213015065 0 302.196078892 94.2212868187 0 380.185781104 260.13113022 377.756904823 359.173392179 355.630301786 249.142194307 160.27004318 186.24688151 253.322831192 398.23973652 256.363676946 139.94470958 283.995081697 164.285202577 0 281.081084739 316.113066325 380.579323314 344.82353539 573.660858505 234.387443432 184.317178088 282.151078215 203.599412447 156.954221265 177.681961245 257.407142197 433.759029456 197.664539362 116.178565051 375.124003282 192.215616735 339.670212162 194.761802587 0 303.140941613 298.522535639 259.872272247 181.217127269 261.848892575 434.07961085 149.484310191 366.411446432 232.266045871 205.14374749 269.731315285 383.64432378 294.403625357 254.884040603 473.947136381 142.368611235 0 135.958462926 287.12117723 328.442813063 347.481748973 170.077399845 387.5448541 311.868863858 133.281349308 194.417269991 452.665515661 176.939783565 374.626919282 196.515439378 253.7539425 338.995645121 287.652392619 393.111562915 295.399190766 394.31304676 153.59724201 0 298.96368446 364.935815972 183.391996053 354.762207843 301.473430139 185.733155873 197.888467273 238.041754298 0 379.535484842 188.188883647 247.534101977 392.61557372 169.478914534 190.638593508 409.605231837 573.129064286 359.279343255 249.300815178 271.710004588 0 311.645598185 234.574433596 326.172936956 324.647778528 261.747214178 306.024829695 0 208.393934892 354.417221067 357.377072857 368.077045728 296.923694788 251.795164161 348.368107081 379.938382912 200.225288051 312.951026113 391.52795884 493.088714607 226.950807288 239.332473056 527.857442863 0 243.819820185 281.082380445 369.863537797 230.450420442 175.90414856 492.075129361 312.457757025 0 311.030329721 382.667752184 389.486264219 443.744729319 369.063794131 200.497564374 165.989412736 185.988529898 272.199763435 369.794418473 329.723110043 0 168.056642907 179.452101302 363.333344502 249.358166016 0 271.20061903 411.66123442 258.769758477 355.305382392 214.472662298 424.355033868 0 206.806421385 189.678063886 451.220710795 0 258.473974099 380.240197168 351.178662986 381.392376154 288.23102823 266.122558126 234.240721809 338.028113771 323.528777504 231.78756735 388.633263705 320.621103711 381.827799196 394.004711212 369.08838335 354.893977299 427.148717788 240.234615239 329.370008691 196.59166987 358.620468557 283.601654226 245.541336149 295.31395711 393.35010182 352.381221364 128.726520221 350.27309649 302.803374468 0 341.055378867 249.290369769 359.765602972 246.588245499 354.021695808 212.683553799 390.17223903 427.274297928 239.991852865 307.062143138 356.189804737 320.023417147 368.103267068 353.488765283 424.701653038 242.633122766 0 338.818145195 454.709449192 202.888496231 266.376504486 372.330611731 565.330375629 310.748228992 0 418.279984823 337.471771326 324.936697164 348.36536037 325.881766453 346.838083798 267.342721865 369.674724693 357.70060661 348.629839303 309.848359142 276.099169225 406.194452551 264.511714689 238.90451191 350.089461776 0 390.961278248 273.132574557 414.893334018 162.293818756 247.492121591 331.258146088 345.816129415 277.347367195 269.043027261 384.174418755 406.292502661 186.214510029 0 521.384192803 362.098603092 340.987727144 0 135.46831619 325.947304106 232.658211598 355.673857102 263.171305616 473.715838882 0 138.439299675 337.435777678 152.74435915 234.795756325 260.776246498 173.331743471 310.805269211 362.656498366 450.395403188 177.727245024 360.096427789 221.620339128 0 349.324967769 212.662891762 188.800936981 364.20202138 302.531661956 387.502932415 285.926213608 306.81129339 297.10060854 252.679213851 252.3992116 275.830148999 378.51323097 0 376.446527179 254.803407818 0 346.107425608 328.412806172 363.365777489 349.067160702 0 344.550617575 385.761333907 292.164994484 408.139716247 329.862925666 333.094240445 470.303935764 301.963158683 0 206.306400808 346.404978575 355.679245791 296.83632308 350.54421094 364.412700153 0 397.69666773 0 355.897330161 147.051046851 471.3181652 422.025785647 355.612599797 449.167987801 471.669875175 337.260559737 114.662749496 421.239520583 272.734176088 206.895174999 305.164076619 219.547418748 298.976817832 384.958953482 192.709212534 0 368.256345659 505.429465354 294.134019396 282.01878855 0 347.3362817 379.737086821 257.108773252 411.918673878 399.008135547 264.437200903 360.219510957 428.514437404 419.627658061 443.703972873 422.178174664 389.820831855 210.072396648 294.373874965 161.83460009 279.961746953 170.473376348 358.703313099 412.715852368 279.356857919 368.9002387 351.317764001 277.434212649 0 310.38723115 164.459607539 223.867384342 324.87200108 176.375882195 0 359.364367356 382.174242839 373.721484205 397.56017535 310.610542534 339.238235375 175.512473115 284.790018213 55 | conv6/sep 267.904562962 56 | fc7 20.4016839339 57 | -------------------------------------------------------------------------------- /classification-dev/resnet18.table: -------------------------------------------------------------------------------- 1 | res2a_branch1_param_0 337.530701467 2 | res2a_branch2a_param_0 478.86380864 3 | res2a_branch2b_param_0 680.018894428 4 | res2b_branch2a_param_0 577.146933522 5 | res2b_branch2b_param_0 425.926449099 6 | res3a_branch1_param_0 229.116841171 7 | res3a_branch2a_param_0 538.17799804 8 | res3a_branch2b_param_0 508.421168649 9 | res3b_branch2a_param_0 760.762128035 10 | res3b_branch2b_param_0 720.429738237 11 | res4a_branch1_param_0 648.576498662 12 | res4a_branch2a_param_0 668.454193599 13 | res4a_branch2b_param_0 548.281064025 14 | res4b_branch2a_param_0 541.042305574 15 | res4b_branch2b_param_0 809.943735936 16 | res5a_branch1_param_0 324.780456345 17 | res5a_branch2a_param_0 872.932131545 18 | res5a_branch2b_param_0 665.028190968 19 | res5b_branch2a_param_0 672.794306006 20 | res5b_branch2b_param_0 708.632196988 21 | res2a_branch1 65.2519381406 22 | res2a_branch2a 65.2519381406 23 | res2a_branch2b 128.527629609 24 | res2b_branch2a 77.1897576464 25 | res2b_branch2b 101.297014977 26 | res3a_branch1 56.8839460552 27 | res3a_branch2a 56.8839460552 28 | res3a_branch2b 98.8722254639 29 | res3b_branch2a 63.931221709 30 | res3b_branch2b 106.839185552 31 | res4a_branch1 80.0833766129 32 | res4a_branch2a 80.0833766129 33 | res4a_branch2b 102.329917766 34 | res4b_branch2a 67.1339920765 35 | res4b_branch2b 107.831406269 36 | res5a_branch1 69.8467121443 37 | res5a_branch2a 69.8467121443 38 | res5a_branch2b 107.474982933 39 | res5b_branch2a 59.8843893023 40 | res5b_branch2b 153.08067073 41 | -------------------------------------------------------------------------------- /classification-dev/squeezenet_v1_1.table: -------------------------------------------------------------------------------- 1 | conv1_param_0 138.066410735 2 | fire2/squeeze1x1_param_0 92.028103407 3 | fire2/expand1x1_param_0 139.5878461 4 | fire2/expand3x3_param_0 170.705731271 5 | fire3/squeeze1x1_param_0 164.533747193 6 | fire3/expand1x1_param_0 130.071935967 7 | fire3/expand3x3_param_0 163.819819809 8 | fire4/squeeze1x1_param_0 116.568016004 9 | fire4/expand1x1_param_0 201.742919491 10 | fire4/expand3x3_param_0 211.399362123 11 | fire5/squeeze1x1_param_0 226.228115652 12 | fire5/expand1x1_param_0 187.774837613 13 | fire5/expand3x3_param_0 164.517905157 14 | fire6/squeeze1x1_param_0 179.423649224 15 | fire6/expand1x1_param_0 266.047646946 16 | fire6/expand3x3_param_0 205.425388157 17 | fire7/squeeze1x1_param_0 208.861314918 18 | fire7/expand1x1_param_0 285.459333589 19 | fire7/expand3x3_param_0 282.445537501 20 | fire8/squeeze1x1_param_0 191.398736522 21 | fire8/expand1x1_param_0 201.295234323 22 | fire8/expand3x3_param_0 240.640160753 23 | fire9/squeeze1x1_param_0 294.368678284 24 | fire9/expand1x1_param_0 244.22330381 25 | fire9/expand3x3_param_0 289.679206026 26 | conv10_param_0 499.107042642 27 | conv1 0.840930858494 28 | fire2/squeeze1x1 0.155449063458 29 | fire2/expand1x1 0.0890687782763 30 | fire2/expand3x3 0.0890687782763 31 | fire3/squeeze1x1 0.136844741617 32 | fire3/expand1x1 0.0876347564071 33 | fire3/expand3x3 0.0876347564071 34 | fire4/squeeze1x1 0.109229065191 35 | fire4/expand1x1 0.0650420251563 36 | fire4/expand3x3 0.0650420251563 37 | fire5/squeeze1x1 0.0931624667139 38 | fire5/expand1x1 0.0775283616794 39 | fire5/expand3x3 0.0775283616794 40 | fire6/squeeze1x1 0.114130987651 41 | fire6/expand1x1 0.0673626595796 42 | fire6/expand3x3 0.0673626595796 43 | fire7/squeeze1x1 0.107491176501 44 | fire7/expand1x1 0.0878467523371 45 | fire7/expand3x3 0.0878467523371 46 | fire8/squeeze1x1 0.144987512392 47 | fire8/expand1x1 0.125130461164 48 | fire8/expand3x3 0.125130461164 49 | fire9/squeeze1x1 0.251220748033 50 | fire9/expand1x1 0.112337101436 51 | fire9/expand3x3 0.112337101436 52 | conv10 0.242779897024 53 | -------------------------------------------------------------------------------- /classification/googlenet_v1.table: -------------------------------------------------------------------------------- 1 | conv2/3x3_reduce_param_0 115.001241494 2 | conv2/3x3_param_0 164.772547138 3 | inception_3a/1x1_param_0 116.860034164 4 | inception_3a/3x3_reduce_param_0 79.1881406724 5 | inception_3a/3x3_param_0 231.357912988 6 | inception_3a/5x5_reduce_param_0 166.105403449 7 | inception_3a/pool_proj_param_0 141.278502738 8 | inception_3b/1x1_param_0 182.756980439 9 | inception_3b/3x3_reduce_param_0 188.983437856 10 | inception_3b/3x3_param_0 210.638443224 11 | inception_3b/5x5_reduce_param_0 191.872441547 12 | inception_3b/pool_proj_param_0 228.95841076 13 | inception_4a/1x1_param_0 230.266832058 14 | inception_4a/3x3_reduce_param_0 213.669944661 15 | inception_4a/3x3_param_0 195.305015429 16 | inception_4a/5x5_reduce_param_0 249.757731161 17 | inception_4a/pool_proj_param_0 188.650596899 18 | inception_4b/1x1_param_0 193.974176486 19 | inception_4b/3x3_reduce_param_0 162.319440983 20 | inception_4b/3x3_param_0 221.563688996 21 | inception_4b/5x5_reduce_param_0 263.254995677 22 | inception_4b/pool_proj_param_0 346.335338297 23 | inception_4c/1x1_param_0 165.722279791 24 | inception_4c/3x3_reduce_param_0 167.280460819 25 | inception_4c/3x3_param_0 314.666309474 26 | inception_4c/5x5_reduce_param_0 307.302805062 27 | inception_4c/pool_proj_param_0 292.606042179 28 | inception_4d/1x1_param_0 168.708545564 29 | inception_4d/3x3_reduce_param_0 241.418758023 30 | inception_4d/3x3_param_0 403.851433463 31 | inception_4d/5x5_reduce_param_0 283.655791452 32 | inception_4d/pool_proj_param_0 315.300139389 33 | inception_4e/1x1_param_0 170.752777459 34 | inception_4e/3x3_reduce_param_0 331.358737323 35 | inception_4e/3x3_param_0 321.119723175 36 | inception_4e/5x5_reduce_param_0 288.571297664 37 | inception_4e/pool_proj_param_0 239.612791756 38 | inception_5a/1x1_param_0 199.742036849 39 | inception_5a/3x3_reduce_param_0 332.047338615 40 | inception_5a/3x3_param_0 815.585545841 41 | inception_5a/5x5_reduce_param_0 267.603533506 42 | inception_5a/pool_proj_param_0 355.915136871 43 | inception_5b/1x1_param_0 356.912443144 44 | inception_5b/3x3_reduce_param_0 372.704808957 45 | inception_5b/3x3_param_0 1142.01213682 46 | inception_5b/5x5_reduce_param_0 500.777341721 47 | inception_5b/pool_proj_param_0 379.110336575 48 | data 0.841264989609 49 | conv1/7x7_s2 0.0693717200182 50 | pool1/3x3_s2 0.0488020289922 51 | pool1/norm1 0.91569411781 52 | conv2/3x3_reduce 0.360708902501 53 | conv2/3x3 0.229959131912 54 | conv2/norm2 0.915694621405 55 | pool2/3x3_s2 0.915694621405 56 | inception_3a/1x1 0.280668467784 57 | inception_3a/3x3_reduce 0.342660584339 58 | inception_3a/3x3 0.1884591056 59 | inception_3a/5x5_reduce 0.319296686606 60 | inception_3a/5x5 0.219277750438 61 | inception_3a/pool 0.915694621405 62 | inception_3a/pool_proj 0.346230151232 63 | inception_3a/output 0.19384928581 64 | inception_3b/1x1 0.310908381737 65 | inception_3b/3x3_reduce 0.26787395878 66 | inception_3b/3x3 0.258738213795 67 | inception_3b/5x5_reduce 0.222254213077 68 | inception_3b/5x5 0.227728164605 69 | inception_3b/pool 0.181090225616 70 | inception_3b/pool_proj 0.273564860946 71 | inception_3b/output 0.254731504282 72 | pool3/3x3_s2 0.22743592442 73 | inception_4a/1x1 0.275513817071 74 | inception_4a/3x3_reduce 0.180952518491 75 | inception_4a/3x3 0.167442384439 76 | inception_4a/5x5_reduce 0.252728857773 77 | inception_4a/5x5 0.322264606512 78 | inception_4a/pool 0.209552993324 79 | inception_4a/pool_proj 0.261629427962 80 | inception_4a/output 0.213100687115 81 | inception_4b/1x1 0.368775609067 82 | inception_4b/3x3_reduce 0.32048056387 83 | inception_4b/3x3 0.315307162368 84 | inception_4b/5x5_reduce 0.253669119528 85 | inception_4b/5x5 0.28602235424 86 | inception_4b/pool 0.170602971096 87 | inception_4b/pool_proj 0.299884916196 88 | inception_4b/output 0.323779907614 89 | inception_4c/1x1 0.357664401528 90 | inception_4c/3x3_reduce 0.315512030229 91 | inception_4c/3x3 0.358776180319 92 | inception_4c/5x5_reduce 0.233842108406 93 | inception_4c/5x5 0.271725636152 94 | inception_4c/pool 0.248194210692 95 | inception_4c/pool_proj 0.341840735446 96 | inception_4c/output 0.317656369042 97 | inception_4d/1x1 0.453047938522 98 | inception_4d/3x3_reduce 0.363660825303 99 | inception_4d/3x3 0.404584425289 100 | inception_4d/5x5_reduce 0.451794309064 101 | inception_4d/5x5 0.438128348527 102 | inception_4d/pool 0.271293431499 103 | inception_4d/pool_proj 0.312833824861 104 | inception_4d/output 0.404955484738 105 | inception_4e/1x1 0.752109277795 106 | inception_4e/3x3_reduce 0.827684286217 107 | inception_4e/3x3 0.873882837486 108 | inception_4e/5x5_reduce 0.599398039464 109 | inception_4e/5x5 0.679890992678 110 | inception_4e/pool 0.318489186187 111 | inception_4e/pool_proj 0.481199466315 112 | inception_4e/output 0.668244529457 113 | pool4/3x3_s2 0.638785256453 114 | inception_5a/1x1 0.929870396787 115 | inception_5a/3x3_reduce 0.850774742469 116 | inception_5a/3x3 1.03021884011 117 | inception_5a/5x5_reduce 0.605795587852 118 | inception_5a/5x5 0.977405003484 119 | inception_5a/pool 0.516934701527 120 | inception_5a/pool_proj 0.619862655889 121 | inception_5a/output 0.804607978019 122 | inception_5b/1x1 2.30040595809 123 | inception_5b/3x3_reduce 1.36618685229 124 | inception_5b/3x3 2.47522426242 125 | inception_5b/5x5_reduce 1.33396191416 126 | inception_5b/5x5 2.51265275518 127 | inception_5b/pool 0.697104108336 128 | inception_5b/pool_proj 1.95249607277 129 | inception_5b/output 2.29461618741 130 | pool5/7x7_s1 11.7747644054 131 | loss3/classifier 4.27423562243 132 | prob 2024.09338521 133 | -------------------------------------------------------------------------------- /classification/mobilenet_v1.table: -------------------------------------------------------------------------------- 1 | conv1_param_0 156.639840536 2 | conv2_1/dw_param_0 63.6101305164 3 | conv2_1/sep_param_0 100.345310403 4 | conv2_2/dw_param_0 255.897909209 5 | conv2_2/sep_param_0 172.178812333 6 | conv3_1/dw_param_0 98.04308453 7 | conv3_1/sep_param_0 128.391132187 8 | conv3_2/dw_param_0 430.423700928 9 | conv3_2/sep_param_0 120.463218716 10 | conv4_1/dw_param_0 154.132992202 11 | conv4_1/sep_param_0 206.227559089 12 | conv4_2/dw_param_0 478.306385215 13 | conv4_2/sep_param_0 197.963457842 14 | conv5_1/dw_param_0 197.52550451 15 | conv5_1/sep_param_0 199.589322533 16 | conv5_2/dw_param_0 194.821766281 17 | conv5_2/sep_param_0 233.45947638 18 | conv5_3/dw_param_0 229.165904235 19 | conv5_3/sep_param_0 223.146231084 20 | conv5_4/dw_param_0 224.154830587 21 | conv5_4/sep_param_0 162.72600108 22 | conv5_5/dw_param_0 228.974207031 23 | conv5_5/sep_param_0 204.163206657 24 | conv5_6/dw_param_0 567.534506444 25 | conv5_6/sep_param_0 286.221061478 26 | conv6/dw_param_0 680.700669437 27 | conv6/sep_param_0 317.761344308 28 | fc7_param_0 211.145021716 29 | data 49.4665182589 30 | conv1 123.881372478 31 | conv2_1/dw 45.3636813192 32 | conv2_1/sep 30.7997748552 33 | conv2_2/dw 38.0147631094 34 | conv2_2/sep 60.2702106607 35 | conv3_1/dw 67.2733262142 36 | conv3_1/sep 62.9261644212 37 | conv3_2/dw 45.1314245571 38 | conv3_2/sep 138.024647679 39 | conv4_1/dw 65.6473314213 40 | conv4_1/sep 108.043928207 41 | conv4_2/dw 112.401202149 42 | conv4_2/sep 159.488017485 43 | conv5_1/dw 84.3091798213 44 | conv5_1/sep 142.616080215 45 | conv5_2/dw 100.286656678 46 | conv5_2/sep 134.686887838 47 | conv5_3/dw 92.0042454653 48 | conv5_3/sep 121.712769475 49 | conv5_4/dw 81.7049592398 50 | conv5_4/sep 89.1531290617 51 | conv5_5/dw 72.6746138205 52 | conv5_5/sep 45.7512218993 53 | conv5_6/dw 101.045986428 54 | conv5_6/sep 233.354378924 55 | conv6/dw 262.792182441 56 | conv6/sep 7.71593719037 57 | pool6 18.9272218404 58 | fc7 4.51432416664 59 | prob 2024.09338521 60 | -------------------------------------------------------------------------------- /classification/mobilenet_v2.table: -------------------------------------------------------------------------------- 1 | conv1_param_0 60.2188558073 2 | conv2_1/expand_param_0 55.8176101762 3 | conv2_1/dwise_param_0 25.5708673441 4 | conv2_1/linear_param_0 52.3772630415 5 | conv2_2/expand_param_0 52.7254689848 6 | conv2_2/dwise_param_0 45.1733737274 7 | conv2_2/linear_param_0 58.793166007 8 | conv3_1/expand_param_0 87.5530858887 9 | conv3_1/dwise_param_0 69.2773295102 10 | conv3_1/linear_param_0 115.29750541 11 | conv3_2/expand_param_0 50.2852335416 12 | conv3_2/dwise_param_0 139.487043266 13 | conv3_2/linear_param_0 57.4598324973 14 | conv4_1/expand_param_0 134.480301318 15 | conv4_1/dwise_param_0 105.254386245 16 | conv4_1/linear_param_0 147.605710018 17 | conv4_2/expand_param_0 145.771279553 18 | conv4_2/dwise_param_0 148.163615473 19 | conv4_2/linear_param_0 152.194501514 20 | conv4_3/expand_param_0 75.9588688975 21 | conv4_3/dwise_param_0 62.4585340138 22 | conv4_3/linear_param_0 97.4342156178 23 | conv4_4/expand_param_0 179.172858442 24 | conv4_4/dwise_param_0 110.525903558 25 | conv4_4/linear_param_0 150.068190211 26 | conv4_5/expand_param_0 138.945489411 27 | conv4_5/dwise_param_0 111.319069363 28 | conv4_5/linear_param_0 152.58800725 29 | conv4_6/expand_param_0 158.762713534 30 | conv4_6/dwise_param_0 111.09766961 31 | conv4_6/linear_param_0 179.574745497 32 | conv4_7/expand_param_0 92.3942439692 33 | conv4_7/dwise_param_0 248.639144087 34 | conv4_7/linear_param_0 114.284912978 35 | conv5_1/expand_param_0 191.593147196 36 | conv5_1/dwise_param_0 90.9081008785 37 | conv5_1/linear_param_0 137.991846285 38 | conv5_2/expand_param_0 170.057763117 39 | conv5_2/dwise_param_0 111.175439814 40 | conv5_2/linear_param_0 192.677955238 41 | conv5_3/expand_param_0 126.925299858 42 | conv5_3/dwise_param_0 174.070581907 43 | conv5_3/linear_param_0 166.389346616 44 | conv6_1/expand_param_0 240.314496931 45 | conv6_1/dwise_param_0 112.186361798 46 | conv6_1/linear_param_0 230.894747987 47 | conv6_2/expand_param_0 164.608908199 48 | conv6_2/dwise_param_0 120.456217828 49 | conv6_2/linear_param_0 213.957776328 50 | conv6_3/expand_param_0 173.981050518 51 | conv6_3/dwise_param_0 186.323267401 52 | conv6_3/linear_param_0 222.61971554 53 | conv6_4_param_0 243.727881679 54 | fc7_param_0 98.2844773267 55 | data 49.4665182589 56 | conv1 16.2354491805 57 | conv2_1/expand 32.76966399 58 | conv2_1/dwise 27.6175879788 59 | conv2_1/linear 7.76812379694 60 | conv2_2/expand 29.4626694763 61 | conv2_2/dwise 34.9570460917 62 | conv2_2/linear 8.74349106467 63 | conv3_1/expand 50.7169143538 64 | conv3_1/dwise 29.4474604799 65 | conv3_1/linear 13.9412717953 66 | block_3_1 8.51389709955 67 | conv3_2/expand 28.9841821121 68 | conv3_2/dwise 19.3444376745 69 | conv3_2/linear 10.9854718531 70 | conv4_1/expand 52.5646259321 71 | conv4_1/dwise 48.5363329539 72 | conv4_1/linear 24.1440463371 73 | block_4_1 11.7046203293 74 | conv4_2/expand 55.8114800186 75 | conv4_2/dwise 43.5167919096 76 | conv4_2/linear 30.8240623836 77 | block_4_2 12.187957786 78 | conv4_3/expand 43.2522777195 79 | conv4_3/dwise 30.0484319679 80 | conv4_3/linear 17.80689606 81 | conv4_4/expand 52.6986254451 82 | conv4_4/dwise 41.4216337821 83 | conv4_4/linear 23.4397885168 84 | block_4_4 14.8823339763 85 | conv4_5/expand 65.2352366523 86 | conv4_5/dwise 47.9830740144 87 | conv4_5/linear 21.3130819147 88 | block_4_5 12.7231607054 89 | conv4_6/expand 66.4366676017 90 | conv4_6/dwise 22.8562254311 91 | conv4_6/linear 17.3041488026 92 | block_4_6 10.8611902167 93 | conv4_7/expand 45.3061942685 94 | conv4_7/dwise 43.0394257382 95 | conv4_7/linear 18.4475329168 96 | conv5_1/expand 54.2424813601 97 | conv5_1/dwise 48.6796850811 98 | conv5_1/linear 22.1433049684 99 | block_5_1 14.6440176281 100 | conv5_2/expand 53.0341003905 101 | conv5_2/dwise 48.249404767 102 | conv5_2/linear 19.8143054709 103 | block_5_2 12.4891413121 104 | conv5_3/expand 55.8543004825 105 | conv5_3/dwise 34.6616360388 106 | conv5_3/linear 18.9427690027 107 | conv6_1/expand 60.0146722566 108 | conv6_1/dwise 63.9853732062 109 | conv6_1/linear 30.1576733788 110 | block_6_1 16.7350757225 111 | conv6_2/expand 69.6402971553 112 | conv6_2/dwise 47.1045164155 113 | conv6_2/linear 24.4669307118 114 | block_6_2 14.5868791007 115 | conv6_3/expand 56.567633418 116 | conv6_3/dwise 43.9295450211 117 | conv6_3/linear 27.6411365729 118 | conv6_4 6.28446365517 119 | pool6 23.8564511938 120 | fc7 3.85843452255 121 | prob 2024.09338521 122 | -------------------------------------------------------------------------------- /classification/resnet18.table: -------------------------------------------------------------------------------- 1 | res2a_branch1_param_0 337.530701 2 | res2a_branch2a_param_0 478.863800 3 | res2a_branch2b_param_0 680.018921 4 | res2b_branch2a_param_0 577.146912 5 | res2b_branch2b_param_0 425.926453 6 | res3a_branch1_param_0 229.116837 7 | res3a_branch2a_param_0 538.177979 8 | res3a_branch2b_param_0 508.421173 9 | res3b_branch2a_param_0 760.762146 10 | res3b_branch2b_param_0 720.429749 11 | res4a_branch1_param_0 648.576477 12 | res4a_branch2a_param_0 668.454224 13 | res4a_branch2b_param_0 548.281067 14 | res4b_branch2a_param_0 541.042297 15 | res4b_branch2b_param_0 809.943726 16 | res5a_branch1_param_0 324.780457 17 | res5a_branch2a_param_0 872.932129 18 | res5a_branch2b_param_0 665.028198 19 | res5b_branch2a_param_0 672.794312 20 | res5b_branch2b_param_0 708.632202 21 | data 0.841265 22 | im_info 9.036131 23 | conv1 90.120193 24 | pool1 61.998196 25 | res2a_branch1 67.005920 26 | res2a_branch2a 128.617905 27 | res2a_branch2b 71.024513 28 | res2a 77.116455 29 | res2b_branch2a 101.862236 30 | res2b_branch2b 67.715729 31 | res2b 56.748234 32 | res3a_branch1 97.656456 33 | res3a_branch2a 96.541336 34 | res3a_branch2b 74.473648 35 | res3a 63.983555 36 | res3b_branch2a 103.834999 37 | res3b_branch2b 85.385017 38 | res3b 79.834740 39 | res4a_branch1 207.663864 40 | res4a_branch2a 100.421661 41 | res4a_branch2b 75.790726 42 | res4a 67.032845 43 | res4b_branch2a 107.825226 44 | res4b_branch2b 68.922409 45 | res4b 69.589668 46 | res5a_branch1 170.960693 47 | res5a_branch2a 107.229637 48 | res5a_branch2b 67.182808 49 | res5a 60.661186 50 | res5b_branch2a 140.155838 51 | res5b_branch2b 7.186572 52 | res5b 5.963103 53 | pool5 13.629847 54 | fc1000 6.299836 55 | prob 2024.096802 56 | -------------------------------------------------------------------------------- /classification/resnet50.table: -------------------------------------------------------------------------------- 1 | res2a_branch1_param_0 141.053626564 2 | res2a_branch2a_param_0 177.00453538 3 | res2a_branch2b_param_0 325.607422427 4 | res2a_branch2c_param_0 319.543374242 5 | res2b_branch2a_param_0 427.814468787 6 | res2b_branch2b_param_0 399.010264273 7 | res2b_branch2c_param_0 454.267242734 8 | res2c_branch2a_param_0 481.589693691 9 | res2c_branch2b_param_0 582.690818258 10 | res2c_branch2c_param_0 362.874112339 11 | res3a_branch1_param_0 197.674232142 12 | res3a_branch2a_param_0 380.313406788 13 | res3a_branch2b_param_0 330.993061159 14 | res3a_branch2c_param_0 292.38987651 15 | res3b_branch2a_param_0 652.279867536 16 | res3b_branch2b_param_0 716.63273835 17 | res3b_branch2c_param_0 368.909033993 18 | res3c_branch2a_param_0 344.635534021 19 | res3c_branch2b_param_0 339.894189489 20 | res3c_branch2c_param_0 441.660118825 21 | res3d_branch2a_param_0 366.771190259 22 | res3d_branch2b_param_0 466.550943171 23 | res3d_branch2c_param_0 452.230413746 24 | res4a_branch1_param_0 301.846209204 25 | res4a_branch2a_param_0 372.653704017 26 | res4a_branch2b_param_0 475.131634056 27 | res4a_branch2c_param_0 330.564508832 28 | res4b_branch2a_param_0 538.295164712 29 | res4b_branch2b_param_0 291.091824937 30 | res4b_branch2c_param_0 290.210275758 31 | res4c_branch2a_param_0 328.244902895 32 | res4c_branch2b_param_0 493.033109849 33 | res4c_branch2c_param_0 438.317438006 34 | res4d_branch2a_param_0 429.919895352 35 | res4d_branch2b_param_0 630.46075603 36 | res4d_branch2c_param_0 530.368312512 37 | res4e_branch2a_param_0 414.888796917 38 | res4e_branch2b_param_0 569.274937965 39 | res4e_branch2c_param_0 479.310424942 40 | res4f_branch2a_param_0 385.203060716 41 | res4f_branch2b_param_0 433.882134167 42 | res4f_branch2c_param_0 432.641333052 43 | res5a_branch1_param_0 204.135313894 44 | res5a_branch2a_param_0 383.708394746 45 | res5a_branch2b_param_0 467.771551136 46 | res5a_branch2c_param_0 294.140643034 47 | res5b_branch2a_param_0 220.22876063 48 | res5b_branch2b_param_0 447.398485922 49 | res5b_branch2c_param_0 458.931333573 50 | res5c_branch2a_param_0 246.859857779 51 | res5c_branch2b_param_0 883.427820616 52 | res5c_branch2c_param_0 423.379791709 53 | data 0.841264989609 54 | conv1 7.54564106623 55 | pool1 6.07490554725 56 | res2a_branch1 7.87266592485 57 | res2a_branch2a 14.0622765726 58 | res2a_branch2b 10.4190757011 59 | res2a_branch2c 13.757139798 60 | res2a 10.0108937316 61 | res2b_branch2a 14.6563565744 62 | res2b_branch2b 13.8603285293 63 | res2b_branch2c 16.5975099094 64 | res2b 11.4420802193 65 | res2c_branch2a 15.7060417523 66 | res2c_branch2b 11.1076179907 67 | res2c_branch2c 17.7396694673 68 | res2c 12.0881140958 69 | res3a_branch1 10.5974755649 70 | res3a_branch2a 14.4895486735 71 | res3a_branch2b 16.8606491774 72 | res3a_branch2c 9.97177797744 73 | res3a 8.76377534438 74 | res3b_branch2a 16.7498528177 75 | res3b_branch2b 14.5755972636 76 | res3b_branch2c 19.0499630796 77 | res3b 10.4547905738 78 | res3c_branch2a 11.6995557384 79 | res3c_branch2b 9.30365131844 80 | res3c_branch2c 8.73712445427 81 | res3c 11.4253567388 82 | res3d_branch2a 10.3075014292 83 | res3d_branch2b 11.407183592 84 | res3d_branch2c 21.1580343298 85 | res3d 8.84143855353 86 | res4a_branch1 9.10587072951 87 | res4a_branch2a 11.7291383944 88 | res4a_branch2b 16.1950230241 89 | res4a_branch2c 12.6165420948 90 | res4a 8.05616223359 91 | res4b_branch2a 14.3075145013 92 | res4b_branch2b 16.7805308692 93 | res4b_branch2c 12.4261290735 94 | res4b 9.50205220349 95 | res4c_branch2a 12.4530584053 96 | res4c_branch2b 16.2050781992 97 | res4c_branch2c 14.35841038 98 | res4c 9.31274788414 99 | res4d_branch2a 14.823744214 100 | res4d_branch2b 12.4014606025 101 | res4d_branch2c 7.90965858811 102 | res4d 9.2993756821 103 | res4e_branch2a 13.2706859006 104 | res4e_branch2b 9.66157444691 105 | res4e_branch2c 7.07866943939 106 | res4e 7.93379927193 107 | res4f_branch2a 10.7007074632 108 | res4f_branch2b 7.82334076201 109 | res4f_branch2c 12.2389186837 110 | res4f 7.24950967996 111 | res5a_branch1 4.91449878237 112 | res5a_branch2a 15.3584133984 113 | res5a_branch2b 19.8853002196 114 | res5a_branch2c 8.11961388066 115 | res5a 3.96442227631 116 | res5b_branch2a 19.1678479719 117 | res5b_branch2b 20.5404800139 118 | res5b_branch2c 8.68713250988 119 | res5b 3.91386263163 120 | res5c_branch2a 16.2805117712 121 | res5c_branch2b 15.6144270351 122 | res5c_branch2c 5.10133887023 123 | res5c 2.6114213527 124 | pool5 15.3136346343 125 | fc1000 3.94138546162 126 | prob 2024.09338521 127 | -------------------------------------------------------------------------------- /classification/squeezenet_v1_1.table: -------------------------------------------------------------------------------- 1 | conv1_param_0 138.066410735 2 | fire2/squeeze1x1_param_0 92.028103407 3 | fire2/expand1x1_param_0 139.5878461 4 | fire2/expand3x3_param_0 170.705731271 5 | fire3/squeeze1x1_param_0 164.533747193 6 | fire3/expand1x1_param_0 130.071935967 7 | fire3/expand3x3_param_0 163.819819809 8 | fire4/squeeze1x1_param_0 116.568016004 9 | fire4/expand1x1_param_0 201.742919491 10 | fire4/expand3x3_param_0 211.399362123 11 | fire5/squeeze1x1_param_0 226.228115652 12 | fire5/expand1x1_param_0 187.774837613 13 | fire5/expand3x3_param_0 164.517905157 14 | fire6/squeeze1x1_param_0 179.423649224 15 | fire6/expand1x1_param_0 266.047646946 16 | fire6/expand3x3_param_0 205.425388157 17 | fire7/squeeze1x1_param_0 208.861314918 18 | fire7/expand1x1_param_0 285.459333589 19 | fire7/expand3x3_param_0 282.445537501 20 | fire8/squeeze1x1_param_0 191.398736522 21 | fire8/expand1x1_param_0 201.295234323 22 | fire8/expand3x3_param_0 240.640160753 23 | fire9/squeeze1x1_param_0 294.368678284 24 | fire9/expand1x1_param_0 244.22330381 25 | fire9/expand3x3_param_0 289.679206026 26 | conv10_param_0 499.107042642 27 | data 0.841264989609 28 | conv1 0.295743466455 29 | pool1 0.161700784564 30 | fire2/squeeze1x1 0.0893839724101 31 | fire2/expand1x1 0.128928090041 32 | fire2/expand3x3 0.13760637974 33 | fire2/concat 0.128852335538 34 | fire3/squeeze1x1 0.0971199525337 35 | fire3/expand1x1 0.171423229656 36 | fire3/expand3x3 0.124295011616 37 | fire3/concat 0.133442531118 38 | pool3 0.109397851642 39 | fire4/squeeze1x1 0.0652759203607 40 | fire4/expand1x1 0.134355569726 41 | fire4/expand3x3 0.0824975446534 42 | fire4/concat 0.0859189409192 43 | fire5/squeeze1x1 0.0819945548106 44 | fire5/expand1x1 0.179214876713 45 | fire5/expand3x3 0.111990834974 46 | fire5/concat 0.128052485114 47 | pool5 0.114620756734 48 | fire6/squeeze1x1 0.0650375683754 49 | fire6/expand1x1 0.155568373518 50 | fire6/expand3x3 0.0936851316279 51 | fire6/concat 0.107651530223 52 | fire7/squeeze1x1 0.087563185785 53 | fire7/expand1x1 0.234552836147 54 | fire7/expand3x3 0.11335142531 55 | fire7/concat 0.145018275864 56 | fire8/squeeze1x1 0.125152598469 57 | fire8/expand1x1 0.439600132251 58 | fire8/expand3x3 0.165414202137 59 | fire8/concat 0.251535496931 60 | fire9/squeeze1x1 0.11218823679 61 | fire9/expand1x1 0.309584429155 62 | fire9/expand3x3 0.184434887035 63 | fire9/concat 0.226503484356 64 | conv10 2.4485219509 65 | pool10 3.32336863982 66 | prob 2024.0936265 67 | -------------------------------------------------------------------------------- /detection/mobilenet_v1-ssd-300.table: -------------------------------------------------------------------------------- 1 | conv0_param_0 27.1343312808 2 | conv1/dw_param_0 23.2944376545 3 | conv1_param_0 21.9425684997 4 | conv2/dw_param_0 19.6928318641 5 | conv2_param_0 48.8967015262 6 | conv3/dw_param_0 21.4390361661 7 | conv3_param_0 48.847920493 8 | conv4/dw_param_0 90.5430318409 9 | conv4_param_0 72.4383360454 10 | conv5/dw_param_0 21.3218941238 11 | conv5_param_0 117.769221662 12 | conv6/dw_param_0 37.6473632119 13 | conv6_param_0 119.587535769 14 | conv7/dw_param_0 27.8276505212 15 | conv7_param_0 145.55121178 16 | conv8/dw_param_0 25.3071462785 17 | conv8_param_0 91.199442267 18 | conv9/dw_param_0 30.5917900825 19 | conv9_param_0 68.2897934839 20 | conv10/dw_param_0 32.4263879234 21 | conv10_param_0 83.7429915813 22 | conv11/dw_param_0 22.3105033013 23 | conv11_param_0 140.123323673 24 | conv12/dw_param_0 6.2146890293 25 | conv12_param_0 133.902477286 26 | conv13/dw_param_0 2.69677908927 27 | conv13_param_0 139.46880996 28 | conv14_1_param_0 217.114642238 29 | conv14_2_param_0 135.036341695 30 | conv15_1_param_0 363.399615913 31 | conv15_2_param_0 157.428430349 32 | conv16_1_param_0 283.463748659 33 | conv16_2_param_0 140.088296847 34 | conv17_1_param_0 181.295474825 35 | conv17_2_param_0 98.4925936862 36 | conv11_mbox_loc_param_0 78.3555422285 37 | conv11_mbox_conf_param_0 23.8404454112 38 | conv13_mbox_loc_param_0 137.15521692 39 | conv13_mbox_conf_param_0 20.8723271414 40 | conv14_2_mbox_loc_param_0 148.255007027 41 | conv14_2_mbox_conf_param_0 21.0791147728 42 | conv15_2_mbox_loc_param_0 156.434490254 43 | conv15_2_mbox_conf_param_0 22.8982352066 44 | conv16_2_mbox_loc_param_0 163.664553431 45 | conv16_2_mbox_conf_param_0 28.246942679 46 | conv17_2_mbox_loc_param_0 262.461097438 47 | conv17_2_mbox_conf_param_0 41.0664206634 48 | data 127.033239532 49 | conv0 10.8546122405 50 | conv1/dw 21.9689174791 51 | conv1 3.37544468118 52 | conv2/dw 19.1445310806 53 | conv2 4.71738709348 54 | conv3/dw 4.59820614046 55 | conv3 3.42106427837 56 | conv4/dw 15.5892518815 57 | conv4 8.66156966607 58 | conv5/dw 14.799565296 59 | conv5 9.56802128064 60 | conv6/dw 15.8153665941 61 | conv6 10.5399075223 62 | conv7/dw 13.1339207446 63 | conv7 14.0346975736 64 | conv8/dw 14.7030592328 65 | conv8 12.370708375 66 | conv9/dw 19.6982374561 67 | conv9 8.14260696401 68 | conv10/dw 19.7400047764 69 | conv10 10.5782605512 70 | conv11/dw 23.3206331016 71 | conv11 16.7585329697 72 | conv12/dw 22.8822106706 73 | conv12 14.7530376997 74 | conv13/dw 28.6683361259 75 | conv13 5.46291092014 76 | conv14_1 23.6808060224 77 | conv14_2 26.9081833296 78 | conv15_1 32.6812935783 79 | conv15_2 20.7481958129 80 | conv16_1 39.1085647054 81 | conv16_2 25.8636239906 82 | conv17_1 21.368959569 83 | conv17_2 14.1117994016 84 | conv11_mbox_loc 7.54745150164 85 | conv11_mbox_conf 0.712908001511 86 | conv13_mbox_loc 4.25555573279 87 | conv13_mbox_conf 0.345332609161 88 | conv14_2_mbox_loc 29.7524446794 89 | conv14_2_mbox_conf 1.10564223976 90 | conv15_2_mbox_loc 27.0860761157 91 | conv15_2_mbox_conf 1.14964248488 92 | conv16_2_mbox_loc 36.0152384467 93 | conv16_2_mbox_conf 0.910049147259 94 | conv17_2_mbox_loc 32.4984184045 95 | conv17_2_mbox_conf 0.675903171496 96 | --------------------------------------------------------------------------------