├── COCO_labels.txt ├── LICENSE ├── MobileNetV1.cpp ├── README.md ├── TestOpenCV_TensorFlow.cbp ├── Traffic.jpg ├── ssd_mobilenet_v1_coco_2017_11_17.pbtxt └── ssd_mobilenet_v2_coco_2018_03_29.pbtxt /COCO_labels.txt: -------------------------------------------------------------------------------- 1 | unlabeled 2 | person 3 | bicycle 4 | car 5 | motorcycle 6 | airplane 7 | bus 8 | train 9 | truck 10 | boat 11 | trafficlight 12 | firehydrant 13 | streetsign 14 | stopsign 15 | parkingmeter 16 | bench 17 | bird 18 | cat 19 | dog 20 | horse 21 | sheep 22 | cow 23 | elephant 24 | bear 25 | zebra 26 | giraffe 27 | hat 28 | backpack 29 | umbrella 30 | shoe 31 | eyeglasses 32 | handbag 33 | tie 34 | suitcase 35 | frisbee 36 | skis 37 | snowboard 38 | sportsball 39 | kite 40 | baseballbat 41 | baseballglove 42 | skateboard 43 | surfboard 44 | tennisracket 45 | bottle 46 | plate 47 | wineglass 48 | cup 49 | fork 50 | knife 51 | spoon 52 | bowl 53 | banana 54 | apple 55 | sandwich 56 | orange 57 | broccoli 58 | carrot 59 | hotdog 60 | pizza 61 | donut 62 | cake 63 | chair 64 | couch 65 | pottedplant 66 | bed 67 | mirror 68 | diningtable 69 | window 70 | desk 71 | toilet 72 | door 73 | tv 74 | laptop 75 | mouse 76 | remote 77 | keyboard 78 | cellphone 79 | microwave 80 | oven 81 | toaster 82 | sink 83 | refrigerator 84 | blender 85 | book 86 | clock 87 | vase 88 | scissors 89 | teddybear 90 | hairdrier 91 | toothbrush -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | BSD 3-Clause License 2 | 3 | Copyright (c) 2021, Q-engineering 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 | 1. Redistributions of source code must retain the above copyright notice, this 10 | list of conditions and the following disclaimer. 11 | 12 | 2. 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 | 3. 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 | -------------------------------------------------------------------------------- /MobileNetV1.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | 9 | using namespace cv; 10 | using namespace std; 11 | 12 | const size_t width = 300; 13 | const size_t height = 300; 14 | 15 | dnn::Net net; 16 | std::vector Names; 17 | 18 | static bool getFileContent(std::string fileName) 19 | { 20 | 21 | // Open the File 22 | std::ifstream in(fileName.c_str()); 23 | // Check if object is valid 24 | if(!in.is_open()) return false; 25 | 26 | std::string str; 27 | // Read the next line from File untill it reaches the end. 28 | while (std::getline(in, str)) 29 | { 30 | // Line contains string of length > 0 then save it in vector 31 | if(str.size()>0) Names.push_back(str); 32 | } 33 | // Close The File 34 | in.close(); 35 | return true; 36 | } 37 | 38 | void detect_from_video(Mat &src) 39 | { 40 | Mat blobimg = dnn::blobFromImage(src, 1.0, Size(300, 300), 0.0, true); 41 | 42 | net.setInput(blobimg); 43 | 44 | Mat detection = net.forward("detection_out"); 45 | // cout << detection.size[2]<<" "<< detection.size[3] << endl; 46 | Mat detectionMat(detection.size[2], detection.size[3], CV_32F, detection.ptr()); 47 | 48 | const float confidence_threshold = 0.25; 49 | for(int i=0; i(i, 2); 51 | 52 | if(detect_confidence > confidence_threshold){ 53 | size_t det_index = (size_t)detectionMat.at(i, 1); 54 | float x1 = detectionMat.at(i, 3)*src.cols; 55 | float y1 = detectionMat.at(i, 4)*src.rows; 56 | float x2 = detectionMat.at(i, 5)*src.cols; 57 | float y2 = detectionMat.at(i, 6)*src.rows; 58 | Rect rec((int)x1, (int)y1, (int)(x2 - x1), (int)(y2 - y1)); 59 | rectangle(src,rec, Scalar(0, 0, 255), 1, 8, 0); 60 | putText(src, format("%s", Names[det_index].c_str()), Point(x1, y1-5) ,FONT_HERSHEY_SIMPLEX,0.5, Scalar(0, 0, 255), 1, 8, 0); 61 | } 62 | } 63 | } 64 | 65 | int main(int argc,char ** argv) 66 | { 67 | float f; 68 | float FPS[16]; 69 | int i, Fcnt=0; 70 | Mat frame; 71 | chrono::steady_clock::time_point Tbegin, Tend; 72 | 73 | for(i=0;i<16;i++) FPS[i]=0.0; 74 | 75 | //MobileNetV1 76 | net = dnn::readNetFromTensorflow("frozen_inference_graph_V1.pb","ssd_mobilenet_v1_coco_2017_11_17.pbtxt"); 77 | //MobileNetV2 78 | //net = dnn::readNetFromTensorflow("frozen_inference_graph_V2.pb","ssd_mobilenet_v2_coco_2018_03_29.pbtxt"); 79 | if (net.empty()){ 80 | cout << "init the model net error"; 81 | exit(-1); 82 | } 83 | 84 | // Get the names 85 | bool result = getFileContent("COCO_labels.txt"); 86 | if(!result) 87 | { 88 | cout << "loading labels failed"; 89 | exit(-1); 90 | } 91 | 92 | //cout << "Switched to " << (cv::ocl::useOpenCL() ? "OpenCL enabled" : "CPU") << endl; 93 | //net.setPreferableTarget(DNN_TARGET_OPENCL); 94 | 95 | cout << "Start grabbing, press ESC on Live window to terminate" << endl; 96 | while(1){ 97 | frame=imread("Traffic.jpg"); //need to refresh frame before dnn class detection 98 | 99 | Tbegin = chrono::steady_clock::now(); 100 | 101 | detect_from_video(frame); 102 | 103 | Tend = chrono::steady_clock::now(); 104 | //calculate frame rate 105 | f = chrono::duration_cast (Tend - Tbegin).count(); 106 | if(f>0.0) FPS[((Fcnt++)&0x0F)]=1000.0/f; 107 | for(f=0.0, i=0;i<16;i++){ f+=FPS[i]; } 108 | putText(frame, format("FPS %0.2f", f/16),Point(10,20),FONT_HERSHEY_SIMPLEX,0.6, Scalar(0, 0, 255)); 109 | //show output 110 | imshow("frame", frame); 111 | 112 | char esc = waitKey(5); 113 | if(esc == 27) break; 114 | } 115 | 116 | cout << "Closing the camera" << endl; 117 | destroyAllWindows(); 118 | cout << "Bye!" << endl; 119 | 120 | return 0; 121 | } 122 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # MobileNetV1/V2_SSD for the DNN modul of OpenCV 2 | ![output image]( https://qengineering.eu/images/V1_FPS.png ) 3 | ## A example of OpenCV dnn framework working on a bare Raspberry Pi with TensorFlow models.
4 | [![License](https://img.shields.io/badge/License-BSD%203--Clause-blue.svg)](https://opensource.org/licenses/BSD-3-Clause)

5 | Paper: https://arxiv.org/abs/1611.10012

6 | Special made for a bare Raspberry Pi 4 see [Q-engineering deep learning examples](https://qengineering.eu/deep-learning-examples-on-raspberry-32-64-os.html) 7 | 8 | ------------ 9 | 10 | Training set: COCO
11 | Size: 300x300
12 | Frame rate V1 : 3.19 FPS (RPi 4)
13 | Frame rate V1_0.75: 4.98 FPS (RPi 4)
14 | Frame rate V2 : 2.02 FPS (RPi 4)
15 | Frame rate V2 Lite: 3.86 FPS (RPi 4)
16 | 17 | 18 | ------------ 19 | 20 | ## Dependencies.
21 | To run the application, you have to: 22 | - A raspberry Pi 4 with a 32 or 64-bit operating system. It can be the Raspberry 64-bit OS, or Ubuntu 18.04 / 20.04. [Install 64-bit OS](https://qengineering.eu/install-raspberry-64-os.html)
23 | - OpenCV 64 bit installed. [Install OpenCV 4.5](https://qengineering.eu/install-opencv-4.5-on-raspberry-64-os.html)
24 | - Code::Blocks installed. (```$ sudo apt-get install codeblocks```) 25 | 26 | ------------ 27 | 28 | ## Installing the app. 29 | To extract and run the network in Code::Blocks
30 | $ mkdir *MyDir*
31 | $ cd *MyDir*
32 | $ wget https://github.com/Qengineering/MobileNet_SSD_OpenCV_TensorFlow/archive/refs/heads/master.zip
33 | $ unzip -j master.zip
34 | Remove master.zip and README.md as they are no longer needed.
35 | $ rm master.zip
36 | $ rm README.md

37 | Your *MyDir* folder must now look like this:
38 | Traffic.jpg
39 | COCO_labels.txt
40 | frozen_inference_graph_V1.pb (download this file from: https://drive.google.com/open?id=1sDn1guYV6oj-AeYuC-riGRh4kv9XBTMz )
41 | frozen_inference_graph_V2.pb (download this file from: https://drive.google.com/open?id=1EU6tVcDNLNwv-pbJUXL7wYUchFHxr5fw )
42 | ssd_mobilenet_v1_coco_2017_11_17.pbtxt
43 | ssd_mobilenet_v2_coco_2018_03_29.pbtxt
44 | TestOpenCV_TensorFlow.cpb
45 | MobileNetV1.cpp (can be use for V2 version also)
46 | 47 | ------------ 48 | 49 | ## Running the app. 50 | To run the application load the project file TestOpenCV_TensorFlow.cbp in Code::Blocks. More info or
51 | if you want to connect a camera to the app, follow the instructions at [Hands-On](https://qengineering.eu/deep-learning-examples-on-raspberry-32-64-os.html#HandsOn).

52 | ![output image]( https://qengineering.eu/images/V1_075_FPS.png ) 53 | ![output image]( https://qengineering.eu/images/V2_FPS.png ) 54 | ![output image]( https://qengineering.eu/images/V2_Lite_FPS.png ) 55 | 56 | ------------ 57 | 58 | [![paypal](https://qengineering.eu/images/TipJarSmall4.png)](https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick&hosted_button_id=CPZTM5BB3FCYL) 59 | -------------------------------------------------------------------------------- /TestOpenCV_TensorFlow.cbp: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 45 | 46 | -------------------------------------------------------------------------------- /Traffic.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Qengineering/MobileNet_SSD_OpenCV_TensorFlow/8ba280cb74b24b50f80e8e3f05b66a68efe39b15/Traffic.jpg -------------------------------------------------------------------------------- /ssd_mobilenet_v1_coco_2017_11_17.pbtxt: -------------------------------------------------------------------------------- 1 | node { 2 | name: "image_tensor" 3 | op: "Placeholder" 4 | attr { 5 | key: "dtype" 6 | value { 7 | type: DT_UINT8 8 | } 9 | } 10 | attr { 11 | key: "shape" 12 | value { 13 | shape { 14 | dim { 15 | size: -1 16 | } 17 | dim { 18 | size: -1 19 | } 20 | dim { 21 | size: -1 22 | } 23 | dim { 24 | size: 3 25 | } 26 | } 27 | } 28 | } 29 | } 30 | node { 31 | name: "Preprocessor/mul" 32 | op: "Mul" 33 | input: "image_tensor" 34 | input: "Preprocessor/mul/x" 35 | } 36 | node { 37 | name: "Preprocessor/sub" 38 | op: "Sub" 39 | input: "Preprocessor/mul" 40 | input: "Preprocessor/sub/y" 41 | } 42 | node { 43 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/mul_1" 44 | op: "Conv2D" 45 | input: "Preprocessor/sub" 46 | input: "FeatureExtractor/MobilenetV1/Conv2d_0/weights/read/_104__cf__107" 47 | input: "^FeatureExtractor/Assert/Assert" 48 | attr { 49 | key: "data_format" 50 | value { 51 | s: "NHWC" 52 | } 53 | } 54 | attr { 55 | key: "padding" 56 | value { 57 | s: "SAME" 58 | } 59 | } 60 | attr { 61 | key: "strides" 62 | value { 63 | list { 64 | i: 1 65 | i: 2 66 | i: 2 67 | i: 1 68 | } 69 | } 70 | } 71 | } 72 | node { 73 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/add_1" 74 | op: "Add" 75 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/mul_1" 76 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/sub/_103__cf__106" 77 | } 78 | node { 79 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/Relu6" 80 | op: "Relu6" 81 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/add_1" 82 | } 83 | node { 84 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/depthwise" 85 | op: "DepthwiseConv2dNative" 86 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/Relu6" 87 | input: "FeatureExtractor/MobilenetV1/Conv2d_1_depthwise/depthwise_weights/read/_101__cf__104" 88 | attr { 89 | key: "data_format" 90 | value { 91 | s: "NHWC" 92 | } 93 | } 94 | attr { 95 | key: "padding" 96 | value { 97 | s: "SAME" 98 | } 99 | } 100 | attr { 101 | key: "strides" 102 | value { 103 | list { 104 | i: 1 105 | i: 1 106 | i: 1 107 | i: 1 108 | } 109 | } 110 | } 111 | } 112 | node { 113 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/BatchNorm/batchnorm/mul_1" 114 | op: "Mul" 115 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/depthwise" 116 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/BatchNorm/batchnorm/mul/_99__cf__102" 117 | } 118 | node { 119 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/BatchNorm/batchnorm/add_1" 120 | op: "Add" 121 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/BatchNorm/batchnorm/mul_1" 122 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/BatchNorm/batchnorm/sub/_100__cf__103" 123 | } 124 | node { 125 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/Relu6" 126 | op: "Relu6" 127 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/BatchNorm/batchnorm/add_1" 128 | } 129 | node { 130 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_pointwise/BatchNorm/batchnorm/mul_1" 131 | op: "Conv2D" 132 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_depthwise/Relu6" 133 | input: "FeatureExtractor/MobilenetV1/Conv2d_1_pointwise/weights/read/_98__cf__101" 134 | attr { 135 | key: "data_format" 136 | value { 137 | s: "NHWC" 138 | } 139 | } 140 | attr { 141 | key: "padding" 142 | value { 143 | s: "SAME" 144 | } 145 | } 146 | attr { 147 | key: "strides" 148 | value { 149 | list { 150 | i: 1 151 | i: 1 152 | i: 1 153 | i: 1 154 | } 155 | } 156 | } 157 | } 158 | node { 159 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_pointwise/BatchNorm/batchnorm/add_1" 160 | op: "Add" 161 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_pointwise/BatchNorm/batchnorm/mul_1" 162 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_pointwise/BatchNorm/batchnorm/sub/_97__cf__100" 163 | } 164 | node { 165 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_pointwise/Relu6" 166 | op: "Relu6" 167 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_pointwise/BatchNorm/batchnorm/add_1" 168 | } 169 | node { 170 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/depthwise" 171 | op: "DepthwiseConv2dNative" 172 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_1_pointwise/Relu6" 173 | input: "FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/depthwise_weights/read/_95__cf__98" 174 | attr { 175 | key: "data_format" 176 | value { 177 | s: "NHWC" 178 | } 179 | } 180 | attr { 181 | key: "padding" 182 | value { 183 | s: "SAME" 184 | } 185 | } 186 | attr { 187 | key: "strides" 188 | value { 189 | list { 190 | i: 1 191 | i: 2 192 | i: 2 193 | i: 1 194 | } 195 | } 196 | } 197 | } 198 | node { 199 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/BatchNorm/batchnorm/mul_1" 200 | op: "Mul" 201 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/depthwise" 202 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/BatchNorm/batchnorm/mul/_93__cf__96" 203 | } 204 | node { 205 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/BatchNorm/batchnorm/add_1" 206 | op: "Add" 207 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/BatchNorm/batchnorm/mul_1" 208 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/BatchNorm/batchnorm/sub/_94__cf__97" 209 | } 210 | node { 211 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/Relu6" 212 | op: "Relu6" 213 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/BatchNorm/batchnorm/add_1" 214 | } 215 | node { 216 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_pointwise/BatchNorm/batchnorm/mul_1" 217 | op: "Conv2D" 218 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_depthwise/Relu6" 219 | input: "FeatureExtractor/MobilenetV1/Conv2d_2_pointwise/weights/read/_92__cf__95" 220 | attr { 221 | key: "data_format" 222 | value { 223 | s: "NHWC" 224 | } 225 | } 226 | attr { 227 | key: "padding" 228 | value { 229 | s: "SAME" 230 | } 231 | } 232 | attr { 233 | key: "strides" 234 | value { 235 | list { 236 | i: 1 237 | i: 1 238 | i: 1 239 | i: 1 240 | } 241 | } 242 | } 243 | } 244 | node { 245 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_pointwise/BatchNorm/batchnorm/add_1" 246 | op: "Add" 247 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_pointwise/BatchNorm/batchnorm/mul_1" 248 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_pointwise/BatchNorm/batchnorm/sub/_91__cf__94" 249 | } 250 | node { 251 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_pointwise/Relu6" 252 | op: "Relu6" 253 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_pointwise/BatchNorm/batchnorm/add_1" 254 | } 255 | node { 256 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/depthwise" 257 | op: "DepthwiseConv2dNative" 258 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_2_pointwise/Relu6" 259 | input: "FeatureExtractor/MobilenetV1/Conv2d_3_depthwise/depthwise_weights/read/_89__cf__92" 260 | attr { 261 | key: "data_format" 262 | value { 263 | s: "NHWC" 264 | } 265 | } 266 | attr { 267 | key: "padding" 268 | value { 269 | s: "SAME" 270 | } 271 | } 272 | attr { 273 | key: "strides" 274 | value { 275 | list { 276 | i: 1 277 | i: 1 278 | i: 1 279 | i: 1 280 | } 281 | } 282 | } 283 | } 284 | node { 285 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/BatchNorm/batchnorm/mul_1" 286 | op: "Mul" 287 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/depthwise" 288 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/BatchNorm/batchnorm/mul/_87__cf__90" 289 | } 290 | node { 291 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/BatchNorm/batchnorm/add_1" 292 | op: "Add" 293 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/BatchNorm/batchnorm/mul_1" 294 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/BatchNorm/batchnorm/sub/_88__cf__91" 295 | } 296 | node { 297 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/Relu6" 298 | op: "Relu6" 299 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/BatchNorm/batchnorm/add_1" 300 | } 301 | node { 302 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_pointwise/BatchNorm/batchnorm/mul_1" 303 | op: "Conv2D" 304 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_depthwise/Relu6" 305 | input: "FeatureExtractor/MobilenetV1/Conv2d_3_pointwise/weights/read/_86__cf__89" 306 | attr { 307 | key: "data_format" 308 | value { 309 | s: "NHWC" 310 | } 311 | } 312 | attr { 313 | key: "padding" 314 | value { 315 | s: "SAME" 316 | } 317 | } 318 | attr { 319 | key: "strides" 320 | value { 321 | list { 322 | i: 1 323 | i: 1 324 | i: 1 325 | i: 1 326 | } 327 | } 328 | } 329 | } 330 | node { 331 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_pointwise/BatchNorm/batchnorm/add_1" 332 | op: "Add" 333 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_pointwise/BatchNorm/batchnorm/mul_1" 334 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_pointwise/BatchNorm/batchnorm/sub/_85__cf__88" 335 | } 336 | node { 337 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_pointwise/Relu6" 338 | op: "Relu6" 339 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_pointwise/BatchNorm/batchnorm/add_1" 340 | } 341 | node { 342 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/depthwise" 343 | op: "DepthwiseConv2dNative" 344 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_3_pointwise/Relu6" 345 | input: "FeatureExtractor/MobilenetV1/Conv2d_4_depthwise/depthwise_weights/read/_83__cf__86" 346 | attr { 347 | key: "data_format" 348 | value { 349 | s: "NHWC" 350 | } 351 | } 352 | attr { 353 | key: "padding" 354 | value { 355 | s: "SAME" 356 | } 357 | } 358 | attr { 359 | key: "strides" 360 | value { 361 | list { 362 | i: 1 363 | i: 2 364 | i: 2 365 | i: 1 366 | } 367 | } 368 | } 369 | } 370 | node { 371 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/BatchNorm/batchnorm/mul_1" 372 | op: "Mul" 373 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/depthwise" 374 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/BatchNorm/batchnorm/mul/_81__cf__84" 375 | } 376 | node { 377 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/BatchNorm/batchnorm/add_1" 378 | op: "Add" 379 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/BatchNorm/batchnorm/mul_1" 380 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/BatchNorm/batchnorm/sub/_82__cf__85" 381 | } 382 | node { 383 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/Relu6" 384 | op: "Relu6" 385 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/BatchNorm/batchnorm/add_1" 386 | } 387 | node { 388 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_pointwise/BatchNorm/batchnorm/mul_1" 389 | op: "Conv2D" 390 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_depthwise/Relu6" 391 | input: "FeatureExtractor/MobilenetV1/Conv2d_4_pointwise/weights/read/_80__cf__83" 392 | attr { 393 | key: "data_format" 394 | value { 395 | s: "NHWC" 396 | } 397 | } 398 | attr { 399 | key: "padding" 400 | value { 401 | s: "SAME" 402 | } 403 | } 404 | attr { 405 | key: "strides" 406 | value { 407 | list { 408 | i: 1 409 | i: 1 410 | i: 1 411 | i: 1 412 | } 413 | } 414 | } 415 | } 416 | node { 417 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_pointwise/BatchNorm/batchnorm/add_1" 418 | op: "Add" 419 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_pointwise/BatchNorm/batchnorm/mul_1" 420 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_pointwise/BatchNorm/batchnorm/sub/_79__cf__82" 421 | } 422 | node { 423 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_pointwise/Relu6" 424 | op: "Relu6" 425 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_pointwise/BatchNorm/batchnorm/add_1" 426 | } 427 | node { 428 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/depthwise" 429 | op: "DepthwiseConv2dNative" 430 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_4_pointwise/Relu6" 431 | input: "FeatureExtractor/MobilenetV1/Conv2d_5_depthwise/depthwise_weights/read/_77__cf__80" 432 | attr { 433 | key: "data_format" 434 | value { 435 | s: "NHWC" 436 | } 437 | } 438 | attr { 439 | key: "padding" 440 | value { 441 | s: "SAME" 442 | } 443 | } 444 | attr { 445 | key: "strides" 446 | value { 447 | list { 448 | i: 1 449 | i: 1 450 | i: 1 451 | i: 1 452 | } 453 | } 454 | } 455 | } 456 | node { 457 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/BatchNorm/batchnorm/mul_1" 458 | op: "Mul" 459 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/depthwise" 460 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/BatchNorm/batchnorm/mul/_75__cf__78" 461 | } 462 | node { 463 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/BatchNorm/batchnorm/add_1" 464 | op: "Add" 465 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/BatchNorm/batchnorm/mul_1" 466 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/BatchNorm/batchnorm/sub/_76__cf__79" 467 | } 468 | node { 469 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/Relu6" 470 | op: "Relu6" 471 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/BatchNorm/batchnorm/add_1" 472 | } 473 | node { 474 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_pointwise/BatchNorm/batchnorm/mul_1" 475 | op: "Conv2D" 476 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_depthwise/Relu6" 477 | input: "FeatureExtractor/MobilenetV1/Conv2d_5_pointwise/weights/read/_74__cf__77" 478 | attr { 479 | key: "data_format" 480 | value { 481 | s: "NHWC" 482 | } 483 | } 484 | attr { 485 | key: "padding" 486 | value { 487 | s: "SAME" 488 | } 489 | } 490 | attr { 491 | key: "strides" 492 | value { 493 | list { 494 | i: 1 495 | i: 1 496 | i: 1 497 | i: 1 498 | } 499 | } 500 | } 501 | } 502 | node { 503 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_pointwise/BatchNorm/batchnorm/add_1" 504 | op: "Add" 505 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_pointwise/BatchNorm/batchnorm/mul_1" 506 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_pointwise/BatchNorm/batchnorm/sub/_73__cf__76" 507 | } 508 | node { 509 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_pointwise/Relu6" 510 | op: "Relu6" 511 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_pointwise/BatchNorm/batchnorm/add_1" 512 | } 513 | node { 514 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/depthwise" 515 | op: "DepthwiseConv2dNative" 516 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_5_pointwise/Relu6" 517 | input: "FeatureExtractor/MobilenetV1/Conv2d_6_depthwise/depthwise_weights/read/_71__cf__74" 518 | attr { 519 | key: "data_format" 520 | value { 521 | s: "NHWC" 522 | } 523 | } 524 | attr { 525 | key: "padding" 526 | value { 527 | s: "SAME" 528 | } 529 | } 530 | attr { 531 | key: "strides" 532 | value { 533 | list { 534 | i: 1 535 | i: 2 536 | i: 2 537 | i: 1 538 | } 539 | } 540 | } 541 | } 542 | node { 543 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/BatchNorm/batchnorm/mul_1" 544 | op: "Mul" 545 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/depthwise" 546 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/BatchNorm/batchnorm/mul/_69__cf__72" 547 | } 548 | node { 549 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/BatchNorm/batchnorm/add_1" 550 | op: "Add" 551 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/BatchNorm/batchnorm/mul_1" 552 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/BatchNorm/batchnorm/sub/_70__cf__73" 553 | } 554 | node { 555 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/Relu6" 556 | op: "Relu6" 557 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/BatchNorm/batchnorm/add_1" 558 | } 559 | node { 560 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_pointwise/BatchNorm/batchnorm/mul_1" 561 | op: "Conv2D" 562 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_depthwise/Relu6" 563 | input: "FeatureExtractor/MobilenetV1/Conv2d_6_pointwise/weights/read/_68__cf__71" 564 | attr { 565 | key: "data_format" 566 | value { 567 | s: "NHWC" 568 | } 569 | } 570 | attr { 571 | key: "padding" 572 | value { 573 | s: "SAME" 574 | } 575 | } 576 | attr { 577 | key: "strides" 578 | value { 579 | list { 580 | i: 1 581 | i: 1 582 | i: 1 583 | i: 1 584 | } 585 | } 586 | } 587 | } 588 | node { 589 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_pointwise/BatchNorm/batchnorm/add_1" 590 | op: "Add" 591 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_pointwise/BatchNorm/batchnorm/mul_1" 592 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_pointwise/BatchNorm/batchnorm/sub/_67__cf__70" 593 | } 594 | node { 595 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_pointwise/Relu6" 596 | op: "Relu6" 597 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_pointwise/BatchNorm/batchnorm/add_1" 598 | } 599 | node { 600 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/depthwise" 601 | op: "DepthwiseConv2dNative" 602 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_6_pointwise/Relu6" 603 | input: "FeatureExtractor/MobilenetV1/Conv2d_7_depthwise/depthwise_weights/read/_65__cf__68" 604 | attr { 605 | key: "data_format" 606 | value { 607 | s: "NHWC" 608 | } 609 | } 610 | attr { 611 | key: "padding" 612 | value { 613 | s: "SAME" 614 | } 615 | } 616 | attr { 617 | key: "strides" 618 | value { 619 | list { 620 | i: 1 621 | i: 1 622 | i: 1 623 | i: 1 624 | } 625 | } 626 | } 627 | } 628 | node { 629 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/BatchNorm/batchnorm/mul_1" 630 | op: "Mul" 631 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/depthwise" 632 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/BatchNorm/batchnorm/mul/_63__cf__66" 633 | } 634 | node { 635 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/BatchNorm/batchnorm/add_1" 636 | op: "Add" 637 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/BatchNorm/batchnorm/mul_1" 638 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/BatchNorm/batchnorm/sub/_64__cf__67" 639 | } 640 | node { 641 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/Relu6" 642 | op: "Relu6" 643 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/BatchNorm/batchnorm/add_1" 644 | } 645 | node { 646 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_pointwise/BatchNorm/batchnorm/mul_1" 647 | op: "Conv2D" 648 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_depthwise/Relu6" 649 | input: "FeatureExtractor/MobilenetV1/Conv2d_7_pointwise/weights/read/_62__cf__65" 650 | attr { 651 | key: "data_format" 652 | value { 653 | s: "NHWC" 654 | } 655 | } 656 | attr { 657 | key: "padding" 658 | value { 659 | s: "SAME" 660 | } 661 | } 662 | attr { 663 | key: "strides" 664 | value { 665 | list { 666 | i: 1 667 | i: 1 668 | i: 1 669 | i: 1 670 | } 671 | } 672 | } 673 | } 674 | node { 675 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_pointwise/BatchNorm/batchnorm/add_1" 676 | op: "Add" 677 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_pointwise/BatchNorm/batchnorm/mul_1" 678 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_pointwise/BatchNorm/batchnorm/sub/_61__cf__64" 679 | } 680 | node { 681 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_pointwise/Relu6" 682 | op: "Relu6" 683 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_pointwise/BatchNorm/batchnorm/add_1" 684 | } 685 | node { 686 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/depthwise" 687 | op: "DepthwiseConv2dNative" 688 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_7_pointwise/Relu6" 689 | input: "FeatureExtractor/MobilenetV1/Conv2d_8_depthwise/depthwise_weights/read/_59__cf__62" 690 | attr { 691 | key: "data_format" 692 | value { 693 | s: "NHWC" 694 | } 695 | } 696 | attr { 697 | key: "padding" 698 | value { 699 | s: "SAME" 700 | } 701 | } 702 | attr { 703 | key: "strides" 704 | value { 705 | list { 706 | i: 1 707 | i: 1 708 | i: 1 709 | i: 1 710 | } 711 | } 712 | } 713 | } 714 | node { 715 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/BatchNorm/batchnorm/mul_1" 716 | op: "Mul" 717 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/depthwise" 718 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/BatchNorm/batchnorm/mul/_57__cf__60" 719 | } 720 | node { 721 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/BatchNorm/batchnorm/add_1" 722 | op: "Add" 723 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/BatchNorm/batchnorm/mul_1" 724 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/BatchNorm/batchnorm/sub/_58__cf__61" 725 | } 726 | node { 727 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/Relu6" 728 | op: "Relu6" 729 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/BatchNorm/batchnorm/add_1" 730 | } 731 | node { 732 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_pointwise/BatchNorm/batchnorm/mul_1" 733 | op: "Conv2D" 734 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_depthwise/Relu6" 735 | input: "FeatureExtractor/MobilenetV1/Conv2d_8_pointwise/weights/read/_56__cf__59" 736 | attr { 737 | key: "data_format" 738 | value { 739 | s: "NHWC" 740 | } 741 | } 742 | attr { 743 | key: "padding" 744 | value { 745 | s: "SAME" 746 | } 747 | } 748 | attr { 749 | key: "strides" 750 | value { 751 | list { 752 | i: 1 753 | i: 1 754 | i: 1 755 | i: 1 756 | } 757 | } 758 | } 759 | } 760 | node { 761 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_pointwise/BatchNorm/batchnorm/add_1" 762 | op: "Add" 763 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_pointwise/BatchNorm/batchnorm/mul_1" 764 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_pointwise/BatchNorm/batchnorm/sub/_55__cf__58" 765 | } 766 | node { 767 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_pointwise/Relu6" 768 | op: "Relu6" 769 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_pointwise/BatchNorm/batchnorm/add_1" 770 | } 771 | node { 772 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/depthwise" 773 | op: "DepthwiseConv2dNative" 774 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_8_pointwise/Relu6" 775 | input: "FeatureExtractor/MobilenetV1/Conv2d_9_depthwise/depthwise_weights/read/_53__cf__56" 776 | attr { 777 | key: "data_format" 778 | value { 779 | s: "NHWC" 780 | } 781 | } 782 | attr { 783 | key: "padding" 784 | value { 785 | s: "SAME" 786 | } 787 | } 788 | attr { 789 | key: "strides" 790 | value { 791 | list { 792 | i: 1 793 | i: 1 794 | i: 1 795 | i: 1 796 | } 797 | } 798 | } 799 | } 800 | node { 801 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/BatchNorm/batchnorm/mul_1" 802 | op: "Mul" 803 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/depthwise" 804 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/BatchNorm/batchnorm/mul/_51__cf__54" 805 | } 806 | node { 807 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/BatchNorm/batchnorm/add_1" 808 | op: "Add" 809 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/BatchNorm/batchnorm/mul_1" 810 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/BatchNorm/batchnorm/sub/_52__cf__55" 811 | } 812 | node { 813 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/Relu6" 814 | op: "Relu6" 815 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/BatchNorm/batchnorm/add_1" 816 | } 817 | node { 818 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_pointwise/BatchNorm/batchnorm/mul_1" 819 | op: "Conv2D" 820 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_depthwise/Relu6" 821 | input: "FeatureExtractor/MobilenetV1/Conv2d_9_pointwise/weights/read/_50__cf__53" 822 | attr { 823 | key: "data_format" 824 | value { 825 | s: "NHWC" 826 | } 827 | } 828 | attr { 829 | key: "padding" 830 | value { 831 | s: "SAME" 832 | } 833 | } 834 | attr { 835 | key: "strides" 836 | value { 837 | list { 838 | i: 1 839 | i: 1 840 | i: 1 841 | i: 1 842 | } 843 | } 844 | } 845 | } 846 | node { 847 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_pointwise/BatchNorm/batchnorm/add_1" 848 | op: "Add" 849 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_pointwise/BatchNorm/batchnorm/mul_1" 850 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_pointwise/BatchNorm/batchnorm/sub/_49__cf__52" 851 | } 852 | node { 853 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_pointwise/Relu6" 854 | op: "Relu6" 855 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_pointwise/BatchNorm/batchnorm/add_1" 856 | } 857 | node { 858 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/depthwise" 859 | op: "DepthwiseConv2dNative" 860 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_9_pointwise/Relu6" 861 | input: "FeatureExtractor/MobilenetV1/Conv2d_10_depthwise/depthwise_weights/read/_47__cf__50" 862 | attr { 863 | key: "data_format" 864 | value { 865 | s: "NHWC" 866 | } 867 | } 868 | attr { 869 | key: "padding" 870 | value { 871 | s: "SAME" 872 | } 873 | } 874 | attr { 875 | key: "strides" 876 | value { 877 | list { 878 | i: 1 879 | i: 1 880 | i: 1 881 | i: 1 882 | } 883 | } 884 | } 885 | } 886 | node { 887 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/BatchNorm/batchnorm/mul_1" 888 | op: "Mul" 889 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/depthwise" 890 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/BatchNorm/batchnorm/mul/_45__cf__48" 891 | } 892 | node { 893 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/BatchNorm/batchnorm/add_1" 894 | op: "Add" 895 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/BatchNorm/batchnorm/mul_1" 896 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/BatchNorm/batchnorm/sub/_46__cf__49" 897 | } 898 | node { 899 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/Relu6" 900 | op: "Relu6" 901 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/BatchNorm/batchnorm/add_1" 902 | } 903 | node { 904 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_pointwise/BatchNorm/batchnorm/mul_1" 905 | op: "Conv2D" 906 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_depthwise/Relu6" 907 | input: "FeatureExtractor/MobilenetV1/Conv2d_10_pointwise/weights/read/_44__cf__47" 908 | attr { 909 | key: "data_format" 910 | value { 911 | s: "NHWC" 912 | } 913 | } 914 | attr { 915 | key: "padding" 916 | value { 917 | s: "SAME" 918 | } 919 | } 920 | attr { 921 | key: "strides" 922 | value { 923 | list { 924 | i: 1 925 | i: 1 926 | i: 1 927 | i: 1 928 | } 929 | } 930 | } 931 | } 932 | node { 933 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_pointwise/BatchNorm/batchnorm/add_1" 934 | op: "Add" 935 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_pointwise/BatchNorm/batchnorm/mul_1" 936 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_pointwise/BatchNorm/batchnorm/sub/_43__cf__46" 937 | } 938 | node { 939 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_pointwise/Relu6" 940 | op: "Relu6" 941 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_pointwise/BatchNorm/batchnorm/add_1" 942 | } 943 | node { 944 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/depthwise" 945 | op: "DepthwiseConv2dNative" 946 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_10_pointwise/Relu6" 947 | input: "FeatureExtractor/MobilenetV1/Conv2d_11_depthwise/depthwise_weights/read/_41__cf__44" 948 | attr { 949 | key: "data_format" 950 | value { 951 | s: "NHWC" 952 | } 953 | } 954 | attr { 955 | key: "padding" 956 | value { 957 | s: "SAME" 958 | } 959 | } 960 | attr { 961 | key: "strides" 962 | value { 963 | list { 964 | i: 1 965 | i: 1 966 | i: 1 967 | i: 1 968 | } 969 | } 970 | } 971 | } 972 | node { 973 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/BatchNorm/batchnorm/mul_1" 974 | op: "Mul" 975 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/depthwise" 976 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/BatchNorm/batchnorm/mul/_39__cf__42" 977 | } 978 | node { 979 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/BatchNorm/batchnorm/add_1" 980 | op: "Add" 981 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/BatchNorm/batchnorm/mul_1" 982 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/BatchNorm/batchnorm/sub/_40__cf__43" 983 | } 984 | node { 985 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/Relu6" 986 | op: "Relu6" 987 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/BatchNorm/batchnorm/add_1" 988 | } 989 | node { 990 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/BatchNorm/batchnorm/mul_1" 991 | op: "Conv2D" 992 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_depthwise/Relu6" 993 | input: "FeatureExtractor/MobilenetV1/Conv2d_11_pointwise/weights/read/_38__cf__41" 994 | attr { 995 | key: "data_format" 996 | value { 997 | s: "NHWC" 998 | } 999 | } 1000 | attr { 1001 | key: "padding" 1002 | value { 1003 | s: "SAME" 1004 | } 1005 | } 1006 | attr { 1007 | key: "strides" 1008 | value { 1009 | list { 1010 | i: 1 1011 | i: 1 1012 | i: 1 1013 | i: 1 1014 | } 1015 | } 1016 | } 1017 | } 1018 | node { 1019 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/BatchNorm/batchnorm/add_1" 1020 | op: "Add" 1021 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/BatchNorm/batchnorm/mul_1" 1022 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/BatchNorm/batchnorm/sub/_37__cf__40" 1023 | } 1024 | node { 1025 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/Relu6" 1026 | op: "Relu6" 1027 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/BatchNorm/batchnorm/add_1" 1028 | } 1029 | node { 1030 | name: "BoxPredictor_0/ClassPredictor/Conv2D" 1031 | op: "Conv2D" 1032 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/Relu6" 1033 | input: "BoxPredictor_0/ClassPredictor/weights/read/_178__cf__181" 1034 | attr { 1035 | key: "data_format" 1036 | value { 1037 | s: "NHWC" 1038 | } 1039 | } 1040 | attr { 1041 | key: "padding" 1042 | value { 1043 | s: "SAME" 1044 | } 1045 | } 1046 | attr { 1047 | key: "strides" 1048 | value { 1049 | list { 1050 | i: 1 1051 | i: 1 1052 | i: 1 1053 | i: 1 1054 | } 1055 | } 1056 | } 1057 | } 1058 | node { 1059 | name: "BoxPredictor_0/ClassPredictor/BiasAdd" 1060 | op: "BiasAdd" 1061 | input: "BoxPredictor_0/ClassPredictor/Conv2D" 1062 | input: "BoxPredictor_0/ClassPredictor/biases/read/_177__cf__180" 1063 | attr { 1064 | key: "data_format" 1065 | value { 1066 | s: "NHWC" 1067 | } 1068 | } 1069 | } 1070 | node { 1071 | name: "BoxPredictor_0/BoxEncodingPredictor/Conv2D" 1072 | op: "Conv2D" 1073 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/Relu6" 1074 | input: "BoxPredictor_0/BoxEncodingPredictor/weights/read/_116__cf__119" 1075 | attr { 1076 | key: "data_format" 1077 | value { 1078 | s: "NHWC" 1079 | } 1080 | } 1081 | attr { 1082 | key: "loc_pred_transposed" 1083 | value { 1084 | b: true 1085 | } 1086 | } 1087 | attr { 1088 | key: "padding" 1089 | value { 1090 | s: "SAME" 1091 | } 1092 | } 1093 | attr { 1094 | key: "strides" 1095 | value { 1096 | list { 1097 | i: 1 1098 | i: 1 1099 | i: 1 1100 | i: 1 1101 | } 1102 | } 1103 | } 1104 | } 1105 | node { 1106 | name: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd" 1107 | op: "BiasAdd" 1108 | input: "BoxPredictor_0/BoxEncodingPredictor/Conv2D" 1109 | input: "BoxPredictor_0/BoxEncodingPredictor/biases/read/_115__cf__118" 1110 | attr { 1111 | key: "data_format" 1112 | value { 1113 | s: "NHWC" 1114 | } 1115 | } 1116 | } 1117 | node { 1118 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/depthwise" 1119 | op: "DepthwiseConv2dNative" 1120 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_11_pointwise/Relu6" 1121 | input: "FeatureExtractor/MobilenetV1/Conv2d_12_depthwise/depthwise_weights/read/_35__cf__38" 1122 | attr { 1123 | key: "data_format" 1124 | value { 1125 | s: "NHWC" 1126 | } 1127 | } 1128 | attr { 1129 | key: "padding" 1130 | value { 1131 | s: "SAME" 1132 | } 1133 | } 1134 | attr { 1135 | key: "strides" 1136 | value { 1137 | list { 1138 | i: 1 1139 | i: 2 1140 | i: 2 1141 | i: 1 1142 | } 1143 | } 1144 | } 1145 | } 1146 | node { 1147 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/BatchNorm/batchnorm/mul_1" 1148 | op: "Mul" 1149 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/depthwise" 1150 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/BatchNorm/batchnorm/mul/_33__cf__36" 1151 | } 1152 | node { 1153 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/BatchNorm/batchnorm/add_1" 1154 | op: "Add" 1155 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/BatchNorm/batchnorm/mul_1" 1156 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/BatchNorm/batchnorm/sub/_34__cf__37" 1157 | } 1158 | node { 1159 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/Relu6" 1160 | op: "Relu6" 1161 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/BatchNorm/batchnorm/add_1" 1162 | } 1163 | node { 1164 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_pointwise/BatchNorm/batchnorm/mul_1" 1165 | op: "Conv2D" 1166 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_depthwise/Relu6" 1167 | input: "FeatureExtractor/MobilenetV1/Conv2d_12_pointwise/weights/read/_32__cf__35" 1168 | attr { 1169 | key: "data_format" 1170 | value { 1171 | s: "NHWC" 1172 | } 1173 | } 1174 | attr { 1175 | key: "padding" 1176 | value { 1177 | s: "SAME" 1178 | } 1179 | } 1180 | attr { 1181 | key: "strides" 1182 | value { 1183 | list { 1184 | i: 1 1185 | i: 1 1186 | i: 1 1187 | i: 1 1188 | } 1189 | } 1190 | } 1191 | } 1192 | node { 1193 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_pointwise/BatchNorm/batchnorm/add_1" 1194 | op: "Add" 1195 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_pointwise/BatchNorm/batchnorm/mul_1" 1196 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_pointwise/BatchNorm/batchnorm/sub/_31__cf__34" 1197 | } 1198 | node { 1199 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_pointwise/Relu6" 1200 | op: "Relu6" 1201 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_pointwise/BatchNorm/batchnorm/add_1" 1202 | } 1203 | node { 1204 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/depthwise" 1205 | op: "DepthwiseConv2dNative" 1206 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_12_pointwise/Relu6" 1207 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_depthwise/depthwise_weights/read/_29__cf__32" 1208 | attr { 1209 | key: "data_format" 1210 | value { 1211 | s: "NHWC" 1212 | } 1213 | } 1214 | attr { 1215 | key: "padding" 1216 | value { 1217 | s: "SAME" 1218 | } 1219 | } 1220 | attr { 1221 | key: "strides" 1222 | value { 1223 | list { 1224 | i: 1 1225 | i: 1 1226 | i: 1 1227 | i: 1 1228 | } 1229 | } 1230 | } 1231 | } 1232 | node { 1233 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/BatchNorm/batchnorm/mul_1" 1234 | op: "Mul" 1235 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/depthwise" 1236 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/BatchNorm/batchnorm/mul/_27__cf__30" 1237 | } 1238 | node { 1239 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/BatchNorm/batchnorm/add_1" 1240 | op: "Add" 1241 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/BatchNorm/batchnorm/mul_1" 1242 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/BatchNorm/batchnorm/sub/_28__cf__31" 1243 | } 1244 | node { 1245 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/Relu6" 1246 | op: "Relu6" 1247 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/BatchNorm/batchnorm/add_1" 1248 | } 1249 | node { 1250 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/BatchNorm/batchnorm/mul_1" 1251 | op: "Conv2D" 1252 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_depthwise/Relu6" 1253 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise/weights/read/_26__cf__29" 1254 | attr { 1255 | key: "data_format" 1256 | value { 1257 | s: "NHWC" 1258 | } 1259 | } 1260 | attr { 1261 | key: "padding" 1262 | value { 1263 | s: "SAME" 1264 | } 1265 | } 1266 | attr { 1267 | key: "strides" 1268 | value { 1269 | list { 1270 | i: 1 1271 | i: 1 1272 | i: 1 1273 | i: 1 1274 | } 1275 | } 1276 | } 1277 | } 1278 | node { 1279 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/BatchNorm/batchnorm/add_1" 1280 | op: "Add" 1281 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/BatchNorm/batchnorm/mul_1" 1282 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/BatchNorm/batchnorm/sub/_25__cf__28" 1283 | } 1284 | node { 1285 | name: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/Relu6" 1286 | op: "Relu6" 1287 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/BatchNorm/batchnorm/add_1" 1288 | } 1289 | node { 1290 | name: "BoxPredictor_1/ClassPredictor/Conv2D" 1291 | op: "Conv2D" 1292 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/Relu6" 1293 | input: "BoxPredictor_1/ClassPredictor/weights/read/_176__cf__179" 1294 | attr { 1295 | key: "data_format" 1296 | value { 1297 | s: "NHWC" 1298 | } 1299 | } 1300 | attr { 1301 | key: "padding" 1302 | value { 1303 | s: "SAME" 1304 | } 1305 | } 1306 | attr { 1307 | key: "strides" 1308 | value { 1309 | list { 1310 | i: 1 1311 | i: 1 1312 | i: 1 1313 | i: 1 1314 | } 1315 | } 1316 | } 1317 | } 1318 | node { 1319 | name: "BoxPredictor_1/ClassPredictor/BiasAdd" 1320 | op: "BiasAdd" 1321 | input: "BoxPredictor_1/ClassPredictor/Conv2D" 1322 | input: "BoxPredictor_1/ClassPredictor/biases/read/_175__cf__178" 1323 | attr { 1324 | key: "data_format" 1325 | value { 1326 | s: "NHWC" 1327 | } 1328 | } 1329 | } 1330 | node { 1331 | name: "BoxPredictor_1/BoxEncodingPredictor/Conv2D" 1332 | op: "Conv2D" 1333 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/Relu6" 1334 | input: "BoxPredictor_1/BoxEncodingPredictor/weights/read/_114__cf__117" 1335 | attr { 1336 | key: "data_format" 1337 | value { 1338 | s: "NHWC" 1339 | } 1340 | } 1341 | attr { 1342 | key: "loc_pred_transposed" 1343 | value { 1344 | b: true 1345 | } 1346 | } 1347 | attr { 1348 | key: "padding" 1349 | value { 1350 | s: "SAME" 1351 | } 1352 | } 1353 | attr { 1354 | key: "strides" 1355 | value { 1356 | list { 1357 | i: 1 1358 | i: 1 1359 | i: 1 1360 | i: 1 1361 | } 1362 | } 1363 | } 1364 | } 1365 | node { 1366 | name: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd" 1367 | op: "BiasAdd" 1368 | input: "BoxPredictor_1/BoxEncodingPredictor/Conv2D" 1369 | input: "BoxPredictor_1/BoxEncodingPredictor/biases/read/_113__cf__116" 1370 | attr { 1371 | key: "data_format" 1372 | value { 1373 | s: "NHWC" 1374 | } 1375 | } 1376 | } 1377 | node { 1378 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/mul_1" 1379 | op: "Conv2D" 1380 | input: "FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_13_pointwise/Relu6" 1381 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/weights/read/_23__cf__26" 1382 | attr { 1383 | key: "data_format" 1384 | value { 1385 | s: "NHWC" 1386 | } 1387 | } 1388 | attr { 1389 | key: "padding" 1390 | value { 1391 | s: "SAME" 1392 | } 1393 | } 1394 | attr { 1395 | key: "strides" 1396 | value { 1397 | list { 1398 | i: 1 1399 | i: 1 1400 | i: 1 1401 | i: 1 1402 | } 1403 | } 1404 | } 1405 | } 1406 | node { 1407 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/add_1" 1408 | op: "Add" 1409 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/mul_1" 1410 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/sub/_22__cf__25" 1411 | } 1412 | node { 1413 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/Relu6" 1414 | op: "Relu6" 1415 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/add_1" 1416 | } 1417 | node { 1418 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/mul_1" 1419 | op: "Conv2D" 1420 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_2_1x1_256/Relu6" 1421 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/weights/read/_20__cf__23" 1422 | attr { 1423 | key: "data_format" 1424 | value { 1425 | s: "NHWC" 1426 | } 1427 | } 1428 | attr { 1429 | key: "padding" 1430 | value { 1431 | s: "SAME" 1432 | } 1433 | } 1434 | attr { 1435 | key: "strides" 1436 | value { 1437 | list { 1438 | i: 1 1439 | i: 2 1440 | i: 2 1441 | i: 1 1442 | } 1443 | } 1444 | } 1445 | } 1446 | node { 1447 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/add_1" 1448 | op: "Add" 1449 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/mul_1" 1450 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/sub/_19__cf__22" 1451 | } 1452 | node { 1453 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/Relu6" 1454 | op: "Relu6" 1455 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/add_1" 1456 | } 1457 | node { 1458 | name: "BoxPredictor_2/ClassPredictor/Conv2D" 1459 | op: "Conv2D" 1460 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/Relu6" 1461 | input: "BoxPredictor_2/ClassPredictor/weights/read/_174__cf__177" 1462 | attr { 1463 | key: "data_format" 1464 | value { 1465 | s: "NHWC" 1466 | } 1467 | } 1468 | attr { 1469 | key: "padding" 1470 | value { 1471 | s: "SAME" 1472 | } 1473 | } 1474 | attr { 1475 | key: "strides" 1476 | value { 1477 | list { 1478 | i: 1 1479 | i: 1 1480 | i: 1 1481 | i: 1 1482 | } 1483 | } 1484 | } 1485 | } 1486 | node { 1487 | name: "BoxPredictor_2/ClassPredictor/BiasAdd" 1488 | op: "BiasAdd" 1489 | input: "BoxPredictor_2/ClassPredictor/Conv2D" 1490 | input: "BoxPredictor_2/ClassPredictor/biases/read/_173__cf__176" 1491 | attr { 1492 | key: "data_format" 1493 | value { 1494 | s: "NHWC" 1495 | } 1496 | } 1497 | } 1498 | node { 1499 | name: "BoxPredictor_2/BoxEncodingPredictor/Conv2D" 1500 | op: "Conv2D" 1501 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/Relu6" 1502 | input: "BoxPredictor_2/BoxEncodingPredictor/weights/read/_112__cf__115" 1503 | attr { 1504 | key: "data_format" 1505 | value { 1506 | s: "NHWC" 1507 | } 1508 | } 1509 | attr { 1510 | key: "loc_pred_transposed" 1511 | value { 1512 | b: true 1513 | } 1514 | } 1515 | attr { 1516 | key: "padding" 1517 | value { 1518 | s: "SAME" 1519 | } 1520 | } 1521 | attr { 1522 | key: "strides" 1523 | value { 1524 | list { 1525 | i: 1 1526 | i: 1 1527 | i: 1 1528 | i: 1 1529 | } 1530 | } 1531 | } 1532 | } 1533 | node { 1534 | name: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd" 1535 | op: "BiasAdd" 1536 | input: "BoxPredictor_2/BoxEncodingPredictor/Conv2D" 1537 | input: "BoxPredictor_2/BoxEncodingPredictor/biases/read/_111__cf__114" 1538 | attr { 1539 | key: "data_format" 1540 | value { 1541 | s: "NHWC" 1542 | } 1543 | } 1544 | } 1545 | node { 1546 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/mul_1" 1547 | op: "Conv2D" 1548 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_2_3x3_s2_512/Relu6" 1549 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/weights/read/_17__cf__20" 1550 | attr { 1551 | key: "data_format" 1552 | value { 1553 | s: "NHWC" 1554 | } 1555 | } 1556 | attr { 1557 | key: "padding" 1558 | value { 1559 | s: "SAME" 1560 | } 1561 | } 1562 | attr { 1563 | key: "strides" 1564 | value { 1565 | list { 1566 | i: 1 1567 | i: 1 1568 | i: 1 1569 | i: 1 1570 | } 1571 | } 1572 | } 1573 | } 1574 | node { 1575 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/add_1" 1576 | op: "Add" 1577 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/mul_1" 1578 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/sub/_16__cf__19" 1579 | } 1580 | node { 1581 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/Relu6" 1582 | op: "Relu6" 1583 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/add_1" 1584 | } 1585 | node { 1586 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/mul_1" 1587 | op: "Conv2D" 1588 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_3_1x1_128/Relu6" 1589 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/weights/read/_14__cf__17" 1590 | attr { 1591 | key: "data_format" 1592 | value { 1593 | s: "NHWC" 1594 | } 1595 | } 1596 | attr { 1597 | key: "padding" 1598 | value { 1599 | s: "SAME" 1600 | } 1601 | } 1602 | attr { 1603 | key: "strides" 1604 | value { 1605 | list { 1606 | i: 1 1607 | i: 2 1608 | i: 2 1609 | i: 1 1610 | } 1611 | } 1612 | } 1613 | } 1614 | node { 1615 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/add_1" 1616 | op: "Add" 1617 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/mul_1" 1618 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/sub/_13__cf__16" 1619 | } 1620 | node { 1621 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/Relu6" 1622 | op: "Relu6" 1623 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/add_1" 1624 | } 1625 | node { 1626 | name: "BoxPredictor_3/ClassPredictor/Conv2D" 1627 | op: "Conv2D" 1628 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/Relu6" 1629 | input: "BoxPredictor_3/ClassPredictor/weights/read/_172__cf__175" 1630 | attr { 1631 | key: "data_format" 1632 | value { 1633 | s: "NHWC" 1634 | } 1635 | } 1636 | attr { 1637 | key: "padding" 1638 | value { 1639 | s: "SAME" 1640 | } 1641 | } 1642 | attr { 1643 | key: "strides" 1644 | value { 1645 | list { 1646 | i: 1 1647 | i: 1 1648 | i: 1 1649 | i: 1 1650 | } 1651 | } 1652 | } 1653 | } 1654 | node { 1655 | name: "BoxPredictor_3/ClassPredictor/BiasAdd" 1656 | op: "BiasAdd" 1657 | input: "BoxPredictor_3/ClassPredictor/Conv2D" 1658 | input: "BoxPredictor_3/ClassPredictor/biases/read/_171__cf__174" 1659 | attr { 1660 | key: "data_format" 1661 | value { 1662 | s: "NHWC" 1663 | } 1664 | } 1665 | } 1666 | node { 1667 | name: "BoxPredictor_3/BoxEncodingPredictor/Conv2D" 1668 | op: "Conv2D" 1669 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/Relu6" 1670 | input: "BoxPredictor_3/BoxEncodingPredictor/weights/read/_110__cf__113" 1671 | attr { 1672 | key: "data_format" 1673 | value { 1674 | s: "NHWC" 1675 | } 1676 | } 1677 | attr { 1678 | key: "loc_pred_transposed" 1679 | value { 1680 | b: true 1681 | } 1682 | } 1683 | attr { 1684 | key: "padding" 1685 | value { 1686 | s: "SAME" 1687 | } 1688 | } 1689 | attr { 1690 | key: "strides" 1691 | value { 1692 | list { 1693 | i: 1 1694 | i: 1 1695 | i: 1 1696 | i: 1 1697 | } 1698 | } 1699 | } 1700 | } 1701 | node { 1702 | name: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd" 1703 | op: "BiasAdd" 1704 | input: "BoxPredictor_3/BoxEncodingPredictor/Conv2D" 1705 | input: "BoxPredictor_3/BoxEncodingPredictor/biases/read/_109__cf__112" 1706 | attr { 1707 | key: "data_format" 1708 | value { 1709 | s: "NHWC" 1710 | } 1711 | } 1712 | } 1713 | node { 1714 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/mul_1" 1715 | op: "Conv2D" 1716 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_3_3x3_s2_256/Relu6" 1717 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/weights/read/_11__cf__14" 1718 | attr { 1719 | key: "data_format" 1720 | value { 1721 | s: "NHWC" 1722 | } 1723 | } 1724 | attr { 1725 | key: "padding" 1726 | value { 1727 | s: "SAME" 1728 | } 1729 | } 1730 | attr { 1731 | key: "strides" 1732 | value { 1733 | list { 1734 | i: 1 1735 | i: 1 1736 | i: 1 1737 | i: 1 1738 | } 1739 | } 1740 | } 1741 | } 1742 | node { 1743 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/add_1" 1744 | op: "Add" 1745 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/mul_1" 1746 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/sub/_10__cf__13" 1747 | } 1748 | node { 1749 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/Relu6" 1750 | op: "Relu6" 1751 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/add_1" 1752 | } 1753 | node { 1754 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/mul_1" 1755 | op: "Conv2D" 1756 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_4_1x1_128/Relu6" 1757 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/weights/read/_8__cf__11" 1758 | attr { 1759 | key: "data_format" 1760 | value { 1761 | s: "NHWC" 1762 | } 1763 | } 1764 | attr { 1765 | key: "padding" 1766 | value { 1767 | s: "SAME" 1768 | } 1769 | } 1770 | attr { 1771 | key: "strides" 1772 | value { 1773 | list { 1774 | i: 1 1775 | i: 2 1776 | i: 2 1777 | i: 1 1778 | } 1779 | } 1780 | } 1781 | } 1782 | node { 1783 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/add_1" 1784 | op: "Add" 1785 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/mul_1" 1786 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/sub/_7__cf__10" 1787 | } 1788 | node { 1789 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/Relu6" 1790 | op: "Relu6" 1791 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/add_1" 1792 | } 1793 | node { 1794 | name: "BoxPredictor_4/ClassPredictor/Conv2D" 1795 | op: "Conv2D" 1796 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/Relu6" 1797 | input: "BoxPredictor_4/ClassPredictor/weights/read/_170__cf__173" 1798 | attr { 1799 | key: "data_format" 1800 | value { 1801 | s: "NHWC" 1802 | } 1803 | } 1804 | attr { 1805 | key: "padding" 1806 | value { 1807 | s: "SAME" 1808 | } 1809 | } 1810 | attr { 1811 | key: "strides" 1812 | value { 1813 | list { 1814 | i: 1 1815 | i: 1 1816 | i: 1 1817 | i: 1 1818 | } 1819 | } 1820 | } 1821 | } 1822 | node { 1823 | name: "BoxPredictor_4/ClassPredictor/BiasAdd" 1824 | op: "BiasAdd" 1825 | input: "BoxPredictor_4/ClassPredictor/Conv2D" 1826 | input: "BoxPredictor_4/ClassPredictor/biases/read/_169__cf__172" 1827 | attr { 1828 | key: "data_format" 1829 | value { 1830 | s: "NHWC" 1831 | } 1832 | } 1833 | } 1834 | node { 1835 | name: "BoxPredictor_4/BoxEncodingPredictor/Conv2D" 1836 | op: "Conv2D" 1837 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/Relu6" 1838 | input: "BoxPredictor_4/BoxEncodingPredictor/weights/read/_108__cf__111" 1839 | attr { 1840 | key: "data_format" 1841 | value { 1842 | s: "NHWC" 1843 | } 1844 | } 1845 | attr { 1846 | key: "loc_pred_transposed" 1847 | value { 1848 | b: true 1849 | } 1850 | } 1851 | attr { 1852 | key: "padding" 1853 | value { 1854 | s: "SAME" 1855 | } 1856 | } 1857 | attr { 1858 | key: "strides" 1859 | value { 1860 | list { 1861 | i: 1 1862 | i: 1 1863 | i: 1 1864 | i: 1 1865 | } 1866 | } 1867 | } 1868 | } 1869 | node { 1870 | name: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd" 1871 | op: "BiasAdd" 1872 | input: "BoxPredictor_4/BoxEncodingPredictor/Conv2D" 1873 | input: "BoxPredictor_4/BoxEncodingPredictor/biases/read/_107__cf__110" 1874 | attr { 1875 | key: "data_format" 1876 | value { 1877 | s: "NHWC" 1878 | } 1879 | } 1880 | } 1881 | node { 1882 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/mul_1" 1883 | op: "Conv2D" 1884 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_4_3x3_s2_256/Relu6" 1885 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/weights/read/_5__cf__8" 1886 | attr { 1887 | key: "data_format" 1888 | value { 1889 | s: "NHWC" 1890 | } 1891 | } 1892 | attr { 1893 | key: "padding" 1894 | value { 1895 | s: "SAME" 1896 | } 1897 | } 1898 | attr { 1899 | key: "strides" 1900 | value { 1901 | list { 1902 | i: 1 1903 | i: 1 1904 | i: 1 1905 | i: 1 1906 | } 1907 | } 1908 | } 1909 | } 1910 | node { 1911 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/add_1" 1912 | op: "Add" 1913 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/mul_1" 1914 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/sub/_4__cf__7" 1915 | } 1916 | node { 1917 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/Relu6" 1918 | op: "Relu6" 1919 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/add_1" 1920 | } 1921 | node { 1922 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/mul_1" 1923 | op: "Conv2D" 1924 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_1_Conv2d_5_1x1_64/Relu6" 1925 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/weights/read/_2__cf__5" 1926 | attr { 1927 | key: "data_format" 1928 | value { 1929 | s: "NHWC" 1930 | } 1931 | } 1932 | attr { 1933 | key: "padding" 1934 | value { 1935 | s: "SAME" 1936 | } 1937 | } 1938 | attr { 1939 | key: "strides" 1940 | value { 1941 | list { 1942 | i: 1 1943 | i: 2 1944 | i: 2 1945 | i: 1 1946 | } 1947 | } 1948 | } 1949 | } 1950 | node { 1951 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/add_1" 1952 | op: "Add" 1953 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/mul_1" 1954 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/sub/_1__cf__4" 1955 | } 1956 | node { 1957 | name: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/Relu6" 1958 | op: "Relu6" 1959 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/add_1" 1960 | } 1961 | node { 1962 | name: "BoxPredictor_5/ClassPredictor/Conv2D" 1963 | op: "Conv2D" 1964 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/Relu6" 1965 | input: "BoxPredictor_5/ClassPredictor/weights/read/_168__cf__171" 1966 | attr { 1967 | key: "data_format" 1968 | value { 1969 | s: "NHWC" 1970 | } 1971 | } 1972 | attr { 1973 | key: "padding" 1974 | value { 1975 | s: "SAME" 1976 | } 1977 | } 1978 | attr { 1979 | key: "strides" 1980 | value { 1981 | list { 1982 | i: 1 1983 | i: 1 1984 | i: 1 1985 | i: 1 1986 | } 1987 | } 1988 | } 1989 | } 1990 | node { 1991 | name: "BoxPredictor_5/ClassPredictor/BiasAdd" 1992 | op: "BiasAdd" 1993 | input: "BoxPredictor_5/ClassPredictor/Conv2D" 1994 | input: "BoxPredictor_5/ClassPredictor/biases/read/_167__cf__170" 1995 | attr { 1996 | key: "data_format" 1997 | value { 1998 | s: "NHWC" 1999 | } 2000 | } 2001 | } 2002 | node { 2003 | name: "BoxPredictor_5/BoxEncodingPredictor/Conv2D" 2004 | op: "Conv2D" 2005 | input: "FeatureExtractor/MobilenetV1/Conv2d_13_pointwise_2_Conv2d_5_3x3_s2_128/Relu6" 2006 | input: "BoxPredictor_5/BoxEncodingPredictor/weights/read/_106__cf__109" 2007 | attr { 2008 | key: "data_format" 2009 | value { 2010 | s: "NHWC" 2011 | } 2012 | } 2013 | attr { 2014 | key: "loc_pred_transposed" 2015 | value { 2016 | b: true 2017 | } 2018 | } 2019 | attr { 2020 | key: "padding" 2021 | value { 2022 | s: "SAME" 2023 | } 2024 | } 2025 | attr { 2026 | key: "strides" 2027 | value { 2028 | list { 2029 | i: 1 2030 | i: 1 2031 | i: 1 2032 | i: 1 2033 | } 2034 | } 2035 | } 2036 | } 2037 | node { 2038 | name: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd" 2039 | op: "BiasAdd" 2040 | input: "BoxPredictor_5/BoxEncodingPredictor/Conv2D" 2041 | input: "BoxPredictor_5/BoxEncodingPredictor/biases/read/_105__cf__108" 2042 | attr { 2043 | key: "data_format" 2044 | value { 2045 | s: "NHWC" 2046 | } 2047 | } 2048 | } 2049 | node { 2050 | name: "concat/axis_flatten" 2051 | op: "Const" 2052 | attr { 2053 | key: "value" 2054 | value { 2055 | tensor { 2056 | dtype: DT_INT32 2057 | int_val: -1 2058 | tensor_shape { 2059 | dim { 2060 | size: 1 2061 | } 2062 | } 2063 | } 2064 | } 2065 | } 2066 | } 2067 | node { 2068 | name: "BoxPredictor_0/ClassPredictor/BiasAdd/Flatten" 2069 | op: "Flatten" 2070 | input: "BoxPredictor_0/ClassPredictor/BiasAdd" 2071 | } 2072 | node { 2073 | name: "BoxPredictor_1/ClassPredictor/BiasAdd/Flatten" 2074 | op: "Flatten" 2075 | input: "BoxPredictor_1/ClassPredictor/BiasAdd" 2076 | } 2077 | node { 2078 | name: "BoxPredictor_2/ClassPredictor/BiasAdd/Flatten" 2079 | op: "Flatten" 2080 | input: "BoxPredictor_2/ClassPredictor/BiasAdd" 2081 | } 2082 | node { 2083 | name: "BoxPredictor_3/ClassPredictor/BiasAdd/Flatten" 2084 | op: "Flatten" 2085 | input: "BoxPredictor_3/ClassPredictor/BiasAdd" 2086 | } 2087 | node { 2088 | name: "BoxPredictor_4/ClassPredictor/BiasAdd/Flatten" 2089 | op: "Flatten" 2090 | input: "BoxPredictor_4/ClassPredictor/BiasAdd" 2091 | } 2092 | node { 2093 | name: "BoxPredictor_5/ClassPredictor/BiasAdd/Flatten" 2094 | op: "Flatten" 2095 | input: "BoxPredictor_5/ClassPredictor/BiasAdd" 2096 | } 2097 | node { 2098 | name: "ClassPredictor/concat" 2099 | op: "ConcatV2" 2100 | input: "BoxPredictor_0/ClassPredictor/BiasAdd/Flatten" 2101 | input: "BoxPredictor_1/ClassPredictor/BiasAdd/Flatten" 2102 | input: "BoxPredictor_2/ClassPredictor/BiasAdd/Flatten" 2103 | input: "BoxPredictor_3/ClassPredictor/BiasAdd/Flatten" 2104 | input: "BoxPredictor_4/ClassPredictor/BiasAdd/Flatten" 2105 | input: "BoxPredictor_5/ClassPredictor/BiasAdd/Flatten" 2106 | input: "concat/axis_flatten" 2107 | } 2108 | node { 2109 | name: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd/Flatten" 2110 | op: "Flatten" 2111 | input: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd" 2112 | } 2113 | node { 2114 | name: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd/Flatten" 2115 | op: "Flatten" 2116 | input: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd" 2117 | } 2118 | node { 2119 | name: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd/Flatten" 2120 | op: "Flatten" 2121 | input: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd" 2122 | } 2123 | node { 2124 | name: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd/Flatten" 2125 | op: "Flatten" 2126 | input: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd" 2127 | } 2128 | node { 2129 | name: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd/Flatten" 2130 | op: "Flatten" 2131 | input: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd" 2132 | } 2133 | node { 2134 | name: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd/Flatten" 2135 | op: "Flatten" 2136 | input: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd" 2137 | } 2138 | node { 2139 | name: "BoxEncodingPredictor/concat" 2140 | op: "ConcatV2" 2141 | input: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd/Flatten" 2142 | input: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd/Flatten" 2143 | input: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd/Flatten" 2144 | input: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd/Flatten" 2145 | input: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd/Flatten" 2146 | input: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd/Flatten" 2147 | input: "concat/axis_flatten" 2148 | } 2149 | node { 2150 | name: "PriorBox_0" 2151 | op: "PriorBox" 2152 | input: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd" 2153 | input: "image_tensor" 2154 | attr { 2155 | key: "clip" 2156 | value { 2157 | b: false 2158 | } 2159 | } 2160 | attr { 2161 | key: "flip" 2162 | value { 2163 | b: false 2164 | } 2165 | } 2166 | attr { 2167 | key: "height" 2168 | value { 2169 | tensor { 2170 | dtype: DT_FLOAT 2171 | float_val: 30.0 2172 | float_val: 42.4264068712 2173 | float_val: 84.8528137424 2174 | tensor_shape { 2175 | dim { 2176 | size: 3 2177 | } 2178 | } 2179 | } 2180 | } 2181 | } 2182 | attr { 2183 | key: "variance" 2184 | value { 2185 | tensor { 2186 | dtype: DT_FLOAT 2187 | float_val: 0.1 2188 | float_val: 0.1 2189 | float_val: 0.2 2190 | float_val: 0.2 2191 | tensor_shape { 2192 | dim { 2193 | size: 4 2194 | } 2195 | } 2196 | } 2197 | } 2198 | } 2199 | attr { 2200 | key: "width" 2201 | value { 2202 | tensor { 2203 | dtype: DT_FLOAT 2204 | float_val: 30.0 2205 | float_val: 84.8528137424 2206 | float_val: 42.4264068712 2207 | tensor_shape { 2208 | dim { 2209 | size: 3 2210 | } 2211 | } 2212 | } 2213 | } 2214 | } 2215 | } 2216 | node { 2217 | name: "PriorBox_1" 2218 | op: "PriorBox" 2219 | input: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd" 2220 | input: "image_tensor" 2221 | attr { 2222 | key: "clip" 2223 | value { 2224 | b: false 2225 | } 2226 | } 2227 | attr { 2228 | key: "flip" 2229 | value { 2230 | b: false 2231 | } 2232 | } 2233 | attr { 2234 | key: "height" 2235 | value { 2236 | tensor { 2237 | dtype: DT_FLOAT 2238 | float_val: 105.0 2239 | float_val: 74.2462120246 2240 | float_val: 148.492424049 2241 | float_val: 60.6217782649 2242 | float_val: 181.874428744 2243 | float_val: 125.49900398 2244 | tensor_shape { 2245 | dim { 2246 | size: 6 2247 | } 2248 | } 2249 | } 2250 | } 2251 | } 2252 | attr { 2253 | key: "variance" 2254 | value { 2255 | tensor { 2256 | dtype: DT_FLOAT 2257 | float_val: 0.1 2258 | float_val: 0.1 2259 | float_val: 0.2 2260 | float_val: 0.2 2261 | tensor_shape { 2262 | dim { 2263 | size: 4 2264 | } 2265 | } 2266 | } 2267 | } 2268 | } 2269 | attr { 2270 | key: "width" 2271 | value { 2272 | tensor { 2273 | dtype: DT_FLOAT 2274 | float_val: 105.0 2275 | float_val: 148.492424049 2276 | float_val: 74.2462120246 2277 | float_val: 181.865334795 2278 | float_val: 60.6187471002 2279 | float_val: 125.49900398 2280 | tensor_shape { 2281 | dim { 2282 | size: 6 2283 | } 2284 | } 2285 | } 2286 | } 2287 | } 2288 | } 2289 | node { 2290 | name: "PriorBox_2" 2291 | op: "PriorBox" 2292 | input: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd" 2293 | input: "image_tensor" 2294 | attr { 2295 | key: "clip" 2296 | value { 2297 | b: false 2298 | } 2299 | } 2300 | attr { 2301 | key: "flip" 2302 | value { 2303 | b: false 2304 | } 2305 | } 2306 | attr { 2307 | key: "height" 2308 | value { 2309 | tensor { 2310 | dtype: DT_FLOAT 2311 | float_val: 150.0 2312 | float_val: 106.066017178 2313 | float_val: 212.132034356 2314 | float_val: 86.6025403784 2315 | float_val: 259.820612491 2316 | float_val: 171.026313765 2317 | tensor_shape { 2318 | dim { 2319 | size: 6 2320 | } 2321 | } 2322 | } 2323 | } 2324 | } 2325 | attr { 2326 | key: "variance" 2327 | value { 2328 | tensor { 2329 | dtype: DT_FLOAT 2330 | float_val: 0.1 2331 | float_val: 0.1 2332 | float_val: 0.2 2333 | float_val: 0.2 2334 | tensor_shape { 2335 | dim { 2336 | size: 4 2337 | } 2338 | } 2339 | } 2340 | } 2341 | } 2342 | attr { 2343 | key: "width" 2344 | value { 2345 | tensor { 2346 | dtype: DT_FLOAT 2347 | float_val: 150.0 2348 | float_val: 212.132034356 2349 | float_val: 106.066017178 2350 | float_val: 259.807621135 2351 | float_val: 86.5982101432 2352 | float_val: 171.026313765 2353 | tensor_shape { 2354 | dim { 2355 | size: 6 2356 | } 2357 | } 2358 | } 2359 | } 2360 | } 2361 | } 2362 | node { 2363 | name: "PriorBox_3" 2364 | op: "PriorBox" 2365 | input: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd" 2366 | input: "image_tensor" 2367 | attr { 2368 | key: "clip" 2369 | value { 2370 | b: false 2371 | } 2372 | } 2373 | attr { 2374 | key: "flip" 2375 | value { 2376 | b: false 2377 | } 2378 | } 2379 | attr { 2380 | key: "height" 2381 | value { 2382 | tensor { 2383 | dtype: DT_FLOAT 2384 | float_val: 195.0 2385 | float_val: 137.885822331 2386 | float_val: 275.771644663 2387 | float_val: 112.583302492 2388 | float_val: 337.766796238 2389 | float_val: 216.333076528 2390 | tensor_shape { 2391 | dim { 2392 | size: 6 2393 | } 2394 | } 2395 | } 2396 | } 2397 | } 2398 | attr { 2399 | key: "variance" 2400 | value { 2401 | tensor { 2402 | dtype: DT_FLOAT 2403 | float_val: 0.1 2404 | float_val: 0.1 2405 | float_val: 0.2 2406 | float_val: 0.2 2407 | tensor_shape { 2408 | dim { 2409 | size: 4 2410 | } 2411 | } 2412 | } 2413 | } 2414 | } 2415 | attr { 2416 | key: "width" 2417 | value { 2418 | tensor { 2419 | dtype: DT_FLOAT 2420 | float_val: 195.0 2421 | float_val: 275.771644663 2422 | float_val: 137.885822331 2423 | float_val: 337.749907476 2424 | float_val: 112.577673186 2425 | float_val: 216.333076528 2426 | tensor_shape { 2427 | dim { 2428 | size: 6 2429 | } 2430 | } 2431 | } 2432 | } 2433 | } 2434 | } 2435 | node { 2436 | name: "PriorBox_4" 2437 | op: "PriorBox" 2438 | input: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd" 2439 | input: "image_tensor" 2440 | attr { 2441 | key: "clip" 2442 | value { 2443 | b: false 2444 | } 2445 | } 2446 | attr { 2447 | key: "flip" 2448 | value { 2449 | b: false 2450 | } 2451 | } 2452 | attr { 2453 | key: "height" 2454 | value { 2455 | tensor { 2456 | dtype: DT_FLOAT 2457 | float_val: 240.0 2458 | float_val: 169.705627485 2459 | float_val: 339.41125497 2460 | float_val: 138.564064606 2461 | float_val: 415.712979985 2462 | float_val: 261.533936612 2463 | tensor_shape { 2464 | dim { 2465 | size: 6 2466 | } 2467 | } 2468 | } 2469 | } 2470 | } 2471 | attr { 2472 | key: "variance" 2473 | value { 2474 | tensor { 2475 | dtype: DT_FLOAT 2476 | float_val: 0.1 2477 | float_val: 0.1 2478 | float_val: 0.2 2479 | float_val: 0.2 2480 | tensor_shape { 2481 | dim { 2482 | size: 4 2483 | } 2484 | } 2485 | } 2486 | } 2487 | } 2488 | attr { 2489 | key: "width" 2490 | value { 2491 | tensor { 2492 | dtype: DT_FLOAT 2493 | float_val: 240.0 2494 | float_val: 339.41125497 2495 | float_val: 169.705627485 2496 | float_val: 415.692193817 2497 | float_val: 138.557136229 2498 | float_val: 261.533936612 2499 | tensor_shape { 2500 | dim { 2501 | size: 6 2502 | } 2503 | } 2504 | } 2505 | } 2506 | } 2507 | } 2508 | node { 2509 | name: "PriorBox_5" 2510 | op: "PriorBox" 2511 | input: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd" 2512 | input: "image_tensor" 2513 | attr { 2514 | key: "clip" 2515 | value { 2516 | b: false 2517 | } 2518 | } 2519 | attr { 2520 | key: "flip" 2521 | value { 2522 | b: false 2523 | } 2524 | } 2525 | attr { 2526 | key: "height" 2527 | value { 2528 | tensor { 2529 | dtype: DT_FLOAT 2530 | float_val: 285.0 2531 | float_val: 201.525432638 2532 | float_val: 403.050865276 2533 | float_val: 164.544826719 2534 | float_val: 493.659163732 2535 | float_val: 292.403830344 2536 | tensor_shape { 2537 | dim { 2538 | size: 6 2539 | } 2540 | } 2541 | } 2542 | } 2543 | } 2544 | attr { 2545 | key: "variance" 2546 | value { 2547 | tensor { 2548 | dtype: DT_FLOAT 2549 | float_val: 0.1 2550 | float_val: 0.1 2551 | float_val: 0.2 2552 | float_val: 0.2 2553 | tensor_shape { 2554 | dim { 2555 | size: 4 2556 | } 2557 | } 2558 | } 2559 | } 2560 | } 2561 | attr { 2562 | key: "width" 2563 | value { 2564 | tensor { 2565 | dtype: DT_FLOAT 2566 | float_val: 285.0 2567 | float_val: 403.050865276 2568 | float_val: 201.525432638 2569 | float_val: 493.634480157 2570 | float_val: 164.536599272 2571 | float_val: 292.403830344 2572 | tensor_shape { 2573 | dim { 2574 | size: 6 2575 | } 2576 | } 2577 | } 2578 | } 2579 | } 2580 | } 2581 | node { 2582 | name: "PriorBox/concat" 2583 | op: "ConcatV2" 2584 | input: "PriorBox_0" 2585 | input: "PriorBox_1" 2586 | input: "PriorBox_2" 2587 | input: "PriorBox_3" 2588 | input: "PriorBox_4" 2589 | input: "PriorBox_5" 2590 | input: "concat/axis_flatten" 2591 | } 2592 | node { 2593 | name: "ClassPredictor/concat/reshape/shape" 2594 | op: "Const" 2595 | attr { 2596 | key: "value" 2597 | value { 2598 | tensor { 2599 | dtype: DT_INT32 2600 | int_val: 0 2601 | int_val: -1 2602 | int_val: 91 2603 | tensor_shape { 2604 | dim { 2605 | size: 3 2606 | } 2607 | } 2608 | } 2609 | } 2610 | } 2611 | } 2612 | node { 2613 | name: "ClassPredictor/concat/reshape" 2614 | op: "Reshape" 2615 | input: "ClassPredictor/concat" 2616 | input: "ClassPredictor/concat/reshape/shape" 2617 | } 2618 | node { 2619 | name: "ClassPredictor/concat/sigmoid" 2620 | op: "Sigmoid" 2621 | input: "ClassPredictor/concat/reshape" 2622 | } 2623 | node { 2624 | name: "ClassPredictor/concat/sigmoid/flatten" 2625 | op: "Flatten" 2626 | input: "ClassPredictor/concat/sigmoid" 2627 | } 2628 | node { 2629 | name: "detection_out" 2630 | op: "DetectionOutput" 2631 | input: "BoxEncodingPredictor/concat" 2632 | input: "ClassPredictor/concat/sigmoid/flatten" 2633 | input: "PriorBox/concat" 2634 | attr { 2635 | key: "background_label_id" 2636 | value { 2637 | i: 0 2638 | } 2639 | } 2640 | attr { 2641 | key: "code_type" 2642 | value { 2643 | s: "CENTER_SIZE" 2644 | } 2645 | } 2646 | attr { 2647 | key: "confidence_threshold" 2648 | value { 2649 | f: 0.01 2650 | } 2651 | } 2652 | attr { 2653 | key: "keep_top_k" 2654 | value { 2655 | i: 100 2656 | } 2657 | } 2658 | attr { 2659 | key: "nms_threshold" 2660 | value { 2661 | f: 0.6 2662 | } 2663 | } 2664 | attr { 2665 | key: "num_classes" 2666 | value { 2667 | i: 91 2668 | } 2669 | } 2670 | attr { 2671 | key: "share_location" 2672 | value { 2673 | b: true 2674 | } 2675 | } 2676 | attr { 2677 | key: "top_k" 2678 | value { 2679 | i: 100 2680 | } 2681 | } 2682 | } 2683 | -------------------------------------------------------------------------------- /ssd_mobilenet_v2_coco_2018_03_29.pbtxt: -------------------------------------------------------------------------------- 1 | node { 2 | name: "image_tensor" 3 | op: "Placeholder" 4 | attr { 5 | key: "dtype" 6 | value { 7 | type: DT_UINT8 8 | } 9 | } 10 | attr { 11 | key: "shape" 12 | value { 13 | shape { 14 | dim { 15 | size: -1 16 | } 17 | dim { 18 | size: -1 19 | } 20 | dim { 21 | size: -1 22 | } 23 | dim { 24 | size: 3 25 | } 26 | } 27 | } 28 | } 29 | } 30 | node { 31 | name: "Preprocessor/mul" 32 | op: "Mul" 33 | input: "image_tensor" 34 | input: "Preprocessor/mul/x" 35 | } 36 | node { 37 | name: "Preprocessor/sub" 38 | op: "Sub" 39 | input: "Preprocessor/mul" 40 | input: "Preprocessor/sub/y" 41 | } 42 | node { 43 | name: "FeatureExtractor/MobilenetV2/Conv/Conv2D" 44 | op: "Conv2D" 45 | input: "Preprocessor/sub" 46 | input: "FeatureExtractor/MobilenetV2/Conv/weights" 47 | attr { 48 | key: "dilations" 49 | value { 50 | list { 51 | i: 1 52 | i: 1 53 | i: 1 54 | i: 1 55 | } 56 | } 57 | } 58 | attr { 59 | key: "padding" 60 | value { 61 | s: "SAME" 62 | } 63 | } 64 | attr { 65 | key: "strides" 66 | value { 67 | list { 68 | i: 1 69 | i: 2 70 | i: 2 71 | i: 1 72 | } 73 | } 74 | } 75 | } 76 | node { 77 | name: "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/add_1" 78 | op: "FusedBatchNorm" 79 | input: "FeatureExtractor/MobilenetV2/Conv/Conv2D" 80 | input: "FeatureExtractor/MobilenetV2/Conv/BatchNorm/gamma" 81 | input: "FeatureExtractor/MobilenetV2/Conv/BatchNorm/beta" 82 | input: "FeatureExtractor/MobilenetV2/Conv/BatchNorm/moving_mean" 83 | input: "FeatureExtractor/MobilenetV2/Conv/BatchNorm/moving_variance" 84 | attr { 85 | key: "epsilon" 86 | value { 87 | f: 0.0010000000475 88 | } 89 | } 90 | } 91 | node { 92 | name: "FeatureExtractor/MobilenetV2/Conv/Relu6" 93 | op: "Relu6" 94 | input: "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/add_1" 95 | } 96 | node { 97 | name: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise" 98 | op: "DepthwiseConv2dNative" 99 | input: "FeatureExtractor/MobilenetV2/Conv/Relu6" 100 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise_weights" 101 | attr { 102 | key: "dilations" 103 | value { 104 | list { 105 | i: 1 106 | i: 1 107 | i: 1 108 | i: 1 109 | } 110 | } 111 | } 112 | attr { 113 | key: "padding" 114 | value { 115 | s: "SAME" 116 | } 117 | } 118 | attr { 119 | key: "strides" 120 | value { 121 | list { 122 | i: 1 123 | i: 1 124 | i: 1 125 | i: 1 126 | } 127 | } 128 | } 129 | } 130 | node { 131 | name: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/add_1" 132 | op: "FusedBatchNorm" 133 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise" 134 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/gamma" 135 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/beta" 136 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/moving_mean" 137 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/moving_variance" 138 | attr { 139 | key: "epsilon" 140 | value { 141 | f: 0.0010000000475 142 | } 143 | } 144 | } 145 | node { 146 | name: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/Relu6" 147 | op: "Relu6" 148 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/add_1" 149 | } 150 | node { 151 | name: "FeatureExtractor/MobilenetV2/expanded_conv/project/Conv2D" 152 | op: "Conv2D" 153 | input: "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/Relu6" 154 | input: "FeatureExtractor/MobilenetV2/expanded_conv/project/weights" 155 | attr { 156 | key: "dilations" 157 | value { 158 | list { 159 | i: 1 160 | i: 1 161 | i: 1 162 | i: 1 163 | } 164 | } 165 | } 166 | attr { 167 | key: "padding" 168 | value { 169 | s: "SAME" 170 | } 171 | } 172 | attr { 173 | key: "strides" 174 | value { 175 | list { 176 | i: 1 177 | i: 1 178 | i: 1 179 | i: 1 180 | } 181 | } 182 | } 183 | } 184 | node { 185 | name: "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/add_1" 186 | op: "FusedBatchNorm" 187 | input: "FeatureExtractor/MobilenetV2/expanded_conv/project/Conv2D" 188 | input: "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/gamma" 189 | input: "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/beta" 190 | input: "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/moving_mean" 191 | input: "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/moving_variance" 192 | attr { 193 | key: "epsilon" 194 | value { 195 | f: 0.0010000000475 196 | } 197 | } 198 | } 199 | node { 200 | name: "FeatureExtractor/MobilenetV2/expanded_conv/output" 201 | op: "Identity" 202 | input: "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/add_1" 203 | } 204 | node { 205 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Conv2D" 206 | op: "Conv2D" 207 | input: "FeatureExtractor/MobilenetV2/expanded_conv/output" 208 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/weights" 209 | attr { 210 | key: "dilations" 211 | value { 212 | list { 213 | i: 1 214 | i: 1 215 | i: 1 216 | i: 1 217 | } 218 | } 219 | } 220 | attr { 221 | key: "padding" 222 | value { 223 | s: "SAME" 224 | } 225 | } 226 | attr { 227 | key: "strides" 228 | value { 229 | list { 230 | i: 1 231 | i: 1 232 | i: 1 233 | i: 1 234 | } 235 | } 236 | } 237 | } 238 | node { 239 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/add_1" 240 | op: "FusedBatchNorm" 241 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Conv2D" 242 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/gamma" 243 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/beta" 244 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/moving_mean" 245 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/moving_variance" 246 | attr { 247 | key: "epsilon" 248 | value { 249 | f: 0.0010000000475 250 | } 251 | } 252 | } 253 | node { 254 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Relu6" 255 | op: "Relu6" 256 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/add_1" 257 | } 258 | node { 259 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise" 260 | op: "DepthwiseConv2dNative" 261 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Relu6" 262 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise_weights" 263 | attr { 264 | key: "dilations" 265 | value { 266 | list { 267 | i: 1 268 | i: 1 269 | i: 1 270 | i: 1 271 | } 272 | } 273 | } 274 | attr { 275 | key: "padding" 276 | value { 277 | s: "SAME" 278 | } 279 | } 280 | attr { 281 | key: "strides" 282 | value { 283 | list { 284 | i: 1 285 | i: 2 286 | i: 2 287 | i: 1 288 | } 289 | } 290 | } 291 | } 292 | node { 293 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/add_1" 294 | op: "FusedBatchNorm" 295 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise" 296 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/gamma" 297 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/beta" 298 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/moving_mean" 299 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/moving_variance" 300 | attr { 301 | key: "epsilon" 302 | value { 303 | f: 0.0010000000475 304 | } 305 | } 306 | } 307 | node { 308 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/Relu6" 309 | op: "Relu6" 310 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/add_1" 311 | } 312 | node { 313 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/Conv2D" 314 | op: "Conv2D" 315 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/Relu6" 316 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/weights" 317 | attr { 318 | key: "dilations" 319 | value { 320 | list { 321 | i: 1 322 | i: 1 323 | i: 1 324 | i: 1 325 | } 326 | } 327 | } 328 | attr { 329 | key: "padding" 330 | value { 331 | s: "SAME" 332 | } 333 | } 334 | attr { 335 | key: "strides" 336 | value { 337 | list { 338 | i: 1 339 | i: 1 340 | i: 1 341 | i: 1 342 | } 343 | } 344 | } 345 | } 346 | node { 347 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/add_1" 348 | op: "FusedBatchNorm" 349 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/Conv2D" 350 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/gamma" 351 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/beta" 352 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/moving_mean" 353 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/moving_variance" 354 | attr { 355 | key: "epsilon" 356 | value { 357 | f: 0.0010000000475 358 | } 359 | } 360 | } 361 | node { 362 | name: "FeatureExtractor/MobilenetV2/expanded_conv_1/output" 363 | op: "Identity" 364 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/add_1" 365 | } 366 | node { 367 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Conv2D" 368 | op: "Conv2D" 369 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/output" 370 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/weights" 371 | attr { 372 | key: "dilations" 373 | value { 374 | list { 375 | i: 1 376 | i: 1 377 | i: 1 378 | i: 1 379 | } 380 | } 381 | } 382 | attr { 383 | key: "padding" 384 | value { 385 | s: "SAME" 386 | } 387 | } 388 | attr { 389 | key: "strides" 390 | value { 391 | list { 392 | i: 1 393 | i: 1 394 | i: 1 395 | i: 1 396 | } 397 | } 398 | } 399 | } 400 | node { 401 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/add_1" 402 | op: "FusedBatchNorm" 403 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Conv2D" 404 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/gamma" 405 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/beta" 406 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/moving_mean" 407 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/moving_variance" 408 | attr { 409 | key: "epsilon" 410 | value { 411 | f: 0.0010000000475 412 | } 413 | } 414 | } 415 | node { 416 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Relu6" 417 | op: "Relu6" 418 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/add_1" 419 | } 420 | node { 421 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise" 422 | op: "DepthwiseConv2dNative" 423 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Relu6" 424 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise_weights" 425 | attr { 426 | key: "dilations" 427 | value { 428 | list { 429 | i: 1 430 | i: 1 431 | i: 1 432 | i: 1 433 | } 434 | } 435 | } 436 | attr { 437 | key: "padding" 438 | value { 439 | s: "SAME" 440 | } 441 | } 442 | attr { 443 | key: "strides" 444 | value { 445 | list { 446 | i: 1 447 | i: 1 448 | i: 1 449 | i: 1 450 | } 451 | } 452 | } 453 | } 454 | node { 455 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/add_1" 456 | op: "FusedBatchNorm" 457 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise" 458 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/gamma" 459 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/beta" 460 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/moving_mean" 461 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/moving_variance" 462 | attr { 463 | key: "epsilon" 464 | value { 465 | f: 0.0010000000475 466 | } 467 | } 468 | } 469 | node { 470 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/Relu6" 471 | op: "Relu6" 472 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/add_1" 473 | } 474 | node { 475 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/Conv2D" 476 | op: "Conv2D" 477 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/Relu6" 478 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/weights" 479 | attr { 480 | key: "dilations" 481 | value { 482 | list { 483 | i: 1 484 | i: 1 485 | i: 1 486 | i: 1 487 | } 488 | } 489 | } 490 | attr { 491 | key: "padding" 492 | value { 493 | s: "SAME" 494 | } 495 | } 496 | attr { 497 | key: "strides" 498 | value { 499 | list { 500 | i: 1 501 | i: 1 502 | i: 1 503 | i: 1 504 | } 505 | } 506 | } 507 | } 508 | node { 509 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/batchnorm/add_1" 510 | op: "FusedBatchNorm" 511 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/Conv2D" 512 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/gamma" 513 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/beta" 514 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/moving_mean" 515 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/moving_variance" 516 | attr { 517 | key: "epsilon" 518 | value { 519 | f: 0.0010000000475 520 | } 521 | } 522 | } 523 | node { 524 | name: "FeatureExtractor/MobilenetV2/expanded_conv_2/add" 525 | op: "Add" 526 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/batchnorm/add_1" 527 | input: "FeatureExtractor/MobilenetV2/expanded_conv_1/output" 528 | } 529 | node { 530 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/input" 531 | op: "Identity" 532 | input: "FeatureExtractor/MobilenetV2/expanded_conv_2/add" 533 | } 534 | node { 535 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Conv2D" 536 | op: "Conv2D" 537 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/input" 538 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/weights" 539 | attr { 540 | key: "dilations" 541 | value { 542 | list { 543 | i: 1 544 | i: 1 545 | i: 1 546 | i: 1 547 | } 548 | } 549 | } 550 | attr { 551 | key: "padding" 552 | value { 553 | s: "SAME" 554 | } 555 | } 556 | attr { 557 | key: "strides" 558 | value { 559 | list { 560 | i: 1 561 | i: 1 562 | i: 1 563 | i: 1 564 | } 565 | } 566 | } 567 | } 568 | node { 569 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/add_1" 570 | op: "FusedBatchNorm" 571 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Conv2D" 572 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/gamma" 573 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/beta" 574 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/moving_mean" 575 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/moving_variance" 576 | attr { 577 | key: "epsilon" 578 | value { 579 | f: 0.0010000000475 580 | } 581 | } 582 | } 583 | node { 584 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Relu6" 585 | op: "Relu6" 586 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/add_1" 587 | } 588 | node { 589 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise" 590 | op: "DepthwiseConv2dNative" 591 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Relu6" 592 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise_weights" 593 | attr { 594 | key: "dilations" 595 | value { 596 | list { 597 | i: 1 598 | i: 1 599 | i: 1 600 | i: 1 601 | } 602 | } 603 | } 604 | attr { 605 | key: "padding" 606 | value { 607 | s: "SAME" 608 | } 609 | } 610 | attr { 611 | key: "strides" 612 | value { 613 | list { 614 | i: 1 615 | i: 2 616 | i: 2 617 | i: 1 618 | } 619 | } 620 | } 621 | } 622 | node { 623 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/add_1" 624 | op: "FusedBatchNorm" 625 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise" 626 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/gamma" 627 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/beta" 628 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/moving_mean" 629 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/moving_variance" 630 | attr { 631 | key: "epsilon" 632 | value { 633 | f: 0.0010000000475 634 | } 635 | } 636 | } 637 | node { 638 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/Relu6" 639 | op: "Relu6" 640 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/add_1" 641 | } 642 | node { 643 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/Conv2D" 644 | op: "Conv2D" 645 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/Relu6" 646 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/weights" 647 | attr { 648 | key: "dilations" 649 | value { 650 | list { 651 | i: 1 652 | i: 1 653 | i: 1 654 | i: 1 655 | } 656 | } 657 | } 658 | attr { 659 | key: "padding" 660 | value { 661 | s: "SAME" 662 | } 663 | } 664 | attr { 665 | key: "strides" 666 | value { 667 | list { 668 | i: 1 669 | i: 1 670 | i: 1 671 | i: 1 672 | } 673 | } 674 | } 675 | } 676 | node { 677 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/add_1" 678 | op: "FusedBatchNorm" 679 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/Conv2D" 680 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/gamma" 681 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/beta" 682 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/moving_mean" 683 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/moving_variance" 684 | attr { 685 | key: "epsilon" 686 | value { 687 | f: 0.0010000000475 688 | } 689 | } 690 | } 691 | node { 692 | name: "FeatureExtractor/MobilenetV2/expanded_conv_3/output" 693 | op: "Identity" 694 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/add_1" 695 | } 696 | node { 697 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Conv2D" 698 | op: "Conv2D" 699 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/output" 700 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/weights" 701 | attr { 702 | key: "dilations" 703 | value { 704 | list { 705 | i: 1 706 | i: 1 707 | i: 1 708 | i: 1 709 | } 710 | } 711 | } 712 | attr { 713 | key: "padding" 714 | value { 715 | s: "SAME" 716 | } 717 | } 718 | attr { 719 | key: "strides" 720 | value { 721 | list { 722 | i: 1 723 | i: 1 724 | i: 1 725 | i: 1 726 | } 727 | } 728 | } 729 | } 730 | node { 731 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/add_1" 732 | op: "FusedBatchNorm" 733 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Conv2D" 734 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/gamma" 735 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/beta" 736 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/moving_mean" 737 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/moving_variance" 738 | attr { 739 | key: "epsilon" 740 | value { 741 | f: 0.0010000000475 742 | } 743 | } 744 | } 745 | node { 746 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Relu6" 747 | op: "Relu6" 748 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/add_1" 749 | } 750 | node { 751 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise" 752 | op: "DepthwiseConv2dNative" 753 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Relu6" 754 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise_weights" 755 | attr { 756 | key: "dilations" 757 | value { 758 | list { 759 | i: 1 760 | i: 1 761 | i: 1 762 | i: 1 763 | } 764 | } 765 | } 766 | attr { 767 | key: "padding" 768 | value { 769 | s: "SAME" 770 | } 771 | } 772 | attr { 773 | key: "strides" 774 | value { 775 | list { 776 | i: 1 777 | i: 1 778 | i: 1 779 | i: 1 780 | } 781 | } 782 | } 783 | } 784 | node { 785 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/add_1" 786 | op: "FusedBatchNorm" 787 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise" 788 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/gamma" 789 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/beta" 790 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/moving_mean" 791 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/moving_variance" 792 | attr { 793 | key: "epsilon" 794 | value { 795 | f: 0.0010000000475 796 | } 797 | } 798 | } 799 | node { 800 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/Relu6" 801 | op: "Relu6" 802 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/add_1" 803 | } 804 | node { 805 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/Conv2D" 806 | op: "Conv2D" 807 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/Relu6" 808 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/weights" 809 | attr { 810 | key: "dilations" 811 | value { 812 | list { 813 | i: 1 814 | i: 1 815 | i: 1 816 | i: 1 817 | } 818 | } 819 | } 820 | attr { 821 | key: "padding" 822 | value { 823 | s: "SAME" 824 | } 825 | } 826 | attr { 827 | key: "strides" 828 | value { 829 | list { 830 | i: 1 831 | i: 1 832 | i: 1 833 | i: 1 834 | } 835 | } 836 | } 837 | } 838 | node { 839 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/batchnorm/add_1" 840 | op: "FusedBatchNorm" 841 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/Conv2D" 842 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/gamma" 843 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/beta" 844 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/moving_mean" 845 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/moving_variance" 846 | attr { 847 | key: "epsilon" 848 | value { 849 | f: 0.0010000000475 850 | } 851 | } 852 | } 853 | node { 854 | name: "FeatureExtractor/MobilenetV2/expanded_conv_4/add" 855 | op: "Add" 856 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/batchnorm/add_1" 857 | input: "FeatureExtractor/MobilenetV2/expanded_conv_3/output" 858 | } 859 | node { 860 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/input" 861 | op: "Identity" 862 | input: "FeatureExtractor/MobilenetV2/expanded_conv_4/add" 863 | } 864 | node { 865 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Conv2D" 866 | op: "Conv2D" 867 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/input" 868 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/weights" 869 | attr { 870 | key: "dilations" 871 | value { 872 | list { 873 | i: 1 874 | i: 1 875 | i: 1 876 | i: 1 877 | } 878 | } 879 | } 880 | attr { 881 | key: "padding" 882 | value { 883 | s: "SAME" 884 | } 885 | } 886 | attr { 887 | key: "strides" 888 | value { 889 | list { 890 | i: 1 891 | i: 1 892 | i: 1 893 | i: 1 894 | } 895 | } 896 | } 897 | } 898 | node { 899 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/add_1" 900 | op: "FusedBatchNorm" 901 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Conv2D" 902 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/gamma" 903 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/beta" 904 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/moving_mean" 905 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/moving_variance" 906 | attr { 907 | key: "epsilon" 908 | value { 909 | f: 0.0010000000475 910 | } 911 | } 912 | } 913 | node { 914 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Relu6" 915 | op: "Relu6" 916 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/add_1" 917 | } 918 | node { 919 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise" 920 | op: "DepthwiseConv2dNative" 921 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Relu6" 922 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise_weights" 923 | attr { 924 | key: "dilations" 925 | value { 926 | list { 927 | i: 1 928 | i: 1 929 | i: 1 930 | i: 1 931 | } 932 | } 933 | } 934 | attr { 935 | key: "padding" 936 | value { 937 | s: "SAME" 938 | } 939 | } 940 | attr { 941 | key: "strides" 942 | value { 943 | list { 944 | i: 1 945 | i: 1 946 | i: 1 947 | i: 1 948 | } 949 | } 950 | } 951 | } 952 | node { 953 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/add_1" 954 | op: "FusedBatchNorm" 955 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise" 956 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/gamma" 957 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/beta" 958 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/moving_mean" 959 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/moving_variance" 960 | attr { 961 | key: "epsilon" 962 | value { 963 | f: 0.0010000000475 964 | } 965 | } 966 | } 967 | node { 968 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/Relu6" 969 | op: "Relu6" 970 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/add_1" 971 | } 972 | node { 973 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/Conv2D" 974 | op: "Conv2D" 975 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/Relu6" 976 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/weights" 977 | attr { 978 | key: "dilations" 979 | value { 980 | list { 981 | i: 1 982 | i: 1 983 | i: 1 984 | i: 1 985 | } 986 | } 987 | } 988 | attr { 989 | key: "padding" 990 | value { 991 | s: "SAME" 992 | } 993 | } 994 | attr { 995 | key: "strides" 996 | value { 997 | list { 998 | i: 1 999 | i: 1 1000 | i: 1 1001 | i: 1 1002 | } 1003 | } 1004 | } 1005 | } 1006 | node { 1007 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/batchnorm/add_1" 1008 | op: "FusedBatchNorm" 1009 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/Conv2D" 1010 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/gamma" 1011 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/beta" 1012 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/moving_mean" 1013 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/moving_variance" 1014 | attr { 1015 | key: "epsilon" 1016 | value { 1017 | f: 0.0010000000475 1018 | } 1019 | } 1020 | } 1021 | node { 1022 | name: "FeatureExtractor/MobilenetV2/expanded_conv_5/add" 1023 | op: "Add" 1024 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/batchnorm/add_1" 1025 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/input" 1026 | } 1027 | node { 1028 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/input" 1029 | op: "Identity" 1030 | input: "FeatureExtractor/MobilenetV2/expanded_conv_5/add" 1031 | } 1032 | node { 1033 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Conv2D" 1034 | op: "Conv2D" 1035 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/input" 1036 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/weights" 1037 | attr { 1038 | key: "dilations" 1039 | value { 1040 | list { 1041 | i: 1 1042 | i: 1 1043 | i: 1 1044 | i: 1 1045 | } 1046 | } 1047 | } 1048 | attr { 1049 | key: "padding" 1050 | value { 1051 | s: "SAME" 1052 | } 1053 | } 1054 | attr { 1055 | key: "strides" 1056 | value { 1057 | list { 1058 | i: 1 1059 | i: 1 1060 | i: 1 1061 | i: 1 1062 | } 1063 | } 1064 | } 1065 | } 1066 | node { 1067 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/add_1" 1068 | op: "FusedBatchNorm" 1069 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Conv2D" 1070 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/gamma" 1071 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/beta" 1072 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/moving_mean" 1073 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/moving_variance" 1074 | attr { 1075 | key: "epsilon" 1076 | value { 1077 | f: 0.0010000000475 1078 | } 1079 | } 1080 | } 1081 | node { 1082 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Relu6" 1083 | op: "Relu6" 1084 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/add_1" 1085 | } 1086 | node { 1087 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise" 1088 | op: "DepthwiseConv2dNative" 1089 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Relu6" 1090 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise_weights" 1091 | attr { 1092 | key: "dilations" 1093 | value { 1094 | list { 1095 | i: 1 1096 | i: 1 1097 | i: 1 1098 | i: 1 1099 | } 1100 | } 1101 | } 1102 | attr { 1103 | key: "padding" 1104 | value { 1105 | s: "SAME" 1106 | } 1107 | } 1108 | attr { 1109 | key: "strides" 1110 | value { 1111 | list { 1112 | i: 1 1113 | i: 2 1114 | i: 2 1115 | i: 1 1116 | } 1117 | } 1118 | } 1119 | } 1120 | node { 1121 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/add_1" 1122 | op: "FusedBatchNorm" 1123 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise" 1124 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/gamma" 1125 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/beta" 1126 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/moving_mean" 1127 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/moving_variance" 1128 | attr { 1129 | key: "epsilon" 1130 | value { 1131 | f: 0.0010000000475 1132 | } 1133 | } 1134 | } 1135 | node { 1136 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/Relu6" 1137 | op: "Relu6" 1138 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/add_1" 1139 | } 1140 | node { 1141 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/Conv2D" 1142 | op: "Conv2D" 1143 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/Relu6" 1144 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/weights" 1145 | attr { 1146 | key: "dilations" 1147 | value { 1148 | list { 1149 | i: 1 1150 | i: 1 1151 | i: 1 1152 | i: 1 1153 | } 1154 | } 1155 | } 1156 | attr { 1157 | key: "padding" 1158 | value { 1159 | s: "SAME" 1160 | } 1161 | } 1162 | attr { 1163 | key: "strides" 1164 | value { 1165 | list { 1166 | i: 1 1167 | i: 1 1168 | i: 1 1169 | i: 1 1170 | } 1171 | } 1172 | } 1173 | } 1174 | node { 1175 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/add_1" 1176 | op: "FusedBatchNorm" 1177 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/Conv2D" 1178 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/gamma" 1179 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/beta" 1180 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/moving_mean" 1181 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/moving_variance" 1182 | attr { 1183 | key: "epsilon" 1184 | value { 1185 | f: 0.0010000000475 1186 | } 1187 | } 1188 | } 1189 | node { 1190 | name: "FeatureExtractor/MobilenetV2/expanded_conv_6/output" 1191 | op: "Identity" 1192 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/add_1" 1193 | } 1194 | node { 1195 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Conv2D" 1196 | op: "Conv2D" 1197 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/output" 1198 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/weights" 1199 | attr { 1200 | key: "dilations" 1201 | value { 1202 | list { 1203 | i: 1 1204 | i: 1 1205 | i: 1 1206 | i: 1 1207 | } 1208 | } 1209 | } 1210 | attr { 1211 | key: "padding" 1212 | value { 1213 | s: "SAME" 1214 | } 1215 | } 1216 | attr { 1217 | key: "strides" 1218 | value { 1219 | list { 1220 | i: 1 1221 | i: 1 1222 | i: 1 1223 | i: 1 1224 | } 1225 | } 1226 | } 1227 | } 1228 | node { 1229 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/add_1" 1230 | op: "FusedBatchNorm" 1231 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Conv2D" 1232 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/gamma" 1233 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/beta" 1234 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/moving_mean" 1235 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/moving_variance" 1236 | attr { 1237 | key: "epsilon" 1238 | value { 1239 | f: 0.0010000000475 1240 | } 1241 | } 1242 | } 1243 | node { 1244 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Relu6" 1245 | op: "Relu6" 1246 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/add_1" 1247 | } 1248 | node { 1249 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise" 1250 | op: "DepthwiseConv2dNative" 1251 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Relu6" 1252 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise_weights" 1253 | attr { 1254 | key: "dilations" 1255 | value { 1256 | list { 1257 | i: 1 1258 | i: 1 1259 | i: 1 1260 | i: 1 1261 | } 1262 | } 1263 | } 1264 | attr { 1265 | key: "padding" 1266 | value { 1267 | s: "SAME" 1268 | } 1269 | } 1270 | attr { 1271 | key: "strides" 1272 | value { 1273 | list { 1274 | i: 1 1275 | i: 1 1276 | i: 1 1277 | i: 1 1278 | } 1279 | } 1280 | } 1281 | } 1282 | node { 1283 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/add_1" 1284 | op: "FusedBatchNorm" 1285 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise" 1286 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/gamma" 1287 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/beta" 1288 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/moving_mean" 1289 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/moving_variance" 1290 | attr { 1291 | key: "epsilon" 1292 | value { 1293 | f: 0.0010000000475 1294 | } 1295 | } 1296 | } 1297 | node { 1298 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/Relu6" 1299 | op: "Relu6" 1300 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/add_1" 1301 | } 1302 | node { 1303 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/Conv2D" 1304 | op: "Conv2D" 1305 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/Relu6" 1306 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/weights" 1307 | attr { 1308 | key: "dilations" 1309 | value { 1310 | list { 1311 | i: 1 1312 | i: 1 1313 | i: 1 1314 | i: 1 1315 | } 1316 | } 1317 | } 1318 | attr { 1319 | key: "padding" 1320 | value { 1321 | s: "SAME" 1322 | } 1323 | } 1324 | attr { 1325 | key: "strides" 1326 | value { 1327 | list { 1328 | i: 1 1329 | i: 1 1330 | i: 1 1331 | i: 1 1332 | } 1333 | } 1334 | } 1335 | } 1336 | node { 1337 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/batchnorm/add_1" 1338 | op: "FusedBatchNorm" 1339 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/Conv2D" 1340 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/gamma" 1341 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/beta" 1342 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/moving_mean" 1343 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/moving_variance" 1344 | attr { 1345 | key: "epsilon" 1346 | value { 1347 | f: 0.0010000000475 1348 | } 1349 | } 1350 | } 1351 | node { 1352 | name: "FeatureExtractor/MobilenetV2/expanded_conv_7/add" 1353 | op: "Add" 1354 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/batchnorm/add_1" 1355 | input: "FeatureExtractor/MobilenetV2/expanded_conv_6/output" 1356 | } 1357 | node { 1358 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/input" 1359 | op: "Identity" 1360 | input: "FeatureExtractor/MobilenetV2/expanded_conv_7/add" 1361 | } 1362 | node { 1363 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Conv2D" 1364 | op: "Conv2D" 1365 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/input" 1366 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/weights" 1367 | attr { 1368 | key: "dilations" 1369 | value { 1370 | list { 1371 | i: 1 1372 | i: 1 1373 | i: 1 1374 | i: 1 1375 | } 1376 | } 1377 | } 1378 | attr { 1379 | key: "padding" 1380 | value { 1381 | s: "SAME" 1382 | } 1383 | } 1384 | attr { 1385 | key: "strides" 1386 | value { 1387 | list { 1388 | i: 1 1389 | i: 1 1390 | i: 1 1391 | i: 1 1392 | } 1393 | } 1394 | } 1395 | } 1396 | node { 1397 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/add_1" 1398 | op: "FusedBatchNorm" 1399 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Conv2D" 1400 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/gamma" 1401 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/beta" 1402 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/moving_mean" 1403 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/moving_variance" 1404 | attr { 1405 | key: "epsilon" 1406 | value { 1407 | f: 0.0010000000475 1408 | } 1409 | } 1410 | } 1411 | node { 1412 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Relu6" 1413 | op: "Relu6" 1414 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/add_1" 1415 | } 1416 | node { 1417 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise" 1418 | op: "DepthwiseConv2dNative" 1419 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Relu6" 1420 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise_weights" 1421 | attr { 1422 | key: "dilations" 1423 | value { 1424 | list { 1425 | i: 1 1426 | i: 1 1427 | i: 1 1428 | i: 1 1429 | } 1430 | } 1431 | } 1432 | attr { 1433 | key: "padding" 1434 | value { 1435 | s: "SAME" 1436 | } 1437 | } 1438 | attr { 1439 | key: "strides" 1440 | value { 1441 | list { 1442 | i: 1 1443 | i: 1 1444 | i: 1 1445 | i: 1 1446 | } 1447 | } 1448 | } 1449 | } 1450 | node { 1451 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/add_1" 1452 | op: "FusedBatchNorm" 1453 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise" 1454 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/gamma" 1455 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/beta" 1456 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/moving_mean" 1457 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/moving_variance" 1458 | attr { 1459 | key: "epsilon" 1460 | value { 1461 | f: 0.0010000000475 1462 | } 1463 | } 1464 | } 1465 | node { 1466 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/Relu6" 1467 | op: "Relu6" 1468 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/add_1" 1469 | } 1470 | node { 1471 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/Conv2D" 1472 | op: "Conv2D" 1473 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/Relu6" 1474 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/weights" 1475 | attr { 1476 | key: "dilations" 1477 | value { 1478 | list { 1479 | i: 1 1480 | i: 1 1481 | i: 1 1482 | i: 1 1483 | } 1484 | } 1485 | } 1486 | attr { 1487 | key: "padding" 1488 | value { 1489 | s: "SAME" 1490 | } 1491 | } 1492 | attr { 1493 | key: "strides" 1494 | value { 1495 | list { 1496 | i: 1 1497 | i: 1 1498 | i: 1 1499 | i: 1 1500 | } 1501 | } 1502 | } 1503 | } 1504 | node { 1505 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/batchnorm/add_1" 1506 | op: "FusedBatchNorm" 1507 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/Conv2D" 1508 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/gamma" 1509 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/beta" 1510 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/moving_mean" 1511 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/moving_variance" 1512 | attr { 1513 | key: "epsilon" 1514 | value { 1515 | f: 0.0010000000475 1516 | } 1517 | } 1518 | } 1519 | node { 1520 | name: "FeatureExtractor/MobilenetV2/expanded_conv_8/add" 1521 | op: "Add" 1522 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/batchnorm/add_1" 1523 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/input" 1524 | } 1525 | node { 1526 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/input" 1527 | op: "Identity" 1528 | input: "FeatureExtractor/MobilenetV2/expanded_conv_8/add" 1529 | } 1530 | node { 1531 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Conv2D" 1532 | op: "Conv2D" 1533 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/input" 1534 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/weights" 1535 | attr { 1536 | key: "dilations" 1537 | value { 1538 | list { 1539 | i: 1 1540 | i: 1 1541 | i: 1 1542 | i: 1 1543 | } 1544 | } 1545 | } 1546 | attr { 1547 | key: "padding" 1548 | value { 1549 | s: "SAME" 1550 | } 1551 | } 1552 | attr { 1553 | key: "strides" 1554 | value { 1555 | list { 1556 | i: 1 1557 | i: 1 1558 | i: 1 1559 | i: 1 1560 | } 1561 | } 1562 | } 1563 | } 1564 | node { 1565 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/add_1" 1566 | op: "FusedBatchNorm" 1567 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Conv2D" 1568 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/gamma" 1569 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/beta" 1570 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/moving_mean" 1571 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/moving_variance" 1572 | attr { 1573 | key: "epsilon" 1574 | value { 1575 | f: 0.0010000000475 1576 | } 1577 | } 1578 | } 1579 | node { 1580 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Relu6" 1581 | op: "Relu6" 1582 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/add_1" 1583 | } 1584 | node { 1585 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise" 1586 | op: "DepthwiseConv2dNative" 1587 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Relu6" 1588 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise_weights" 1589 | attr { 1590 | key: "dilations" 1591 | value { 1592 | list { 1593 | i: 1 1594 | i: 1 1595 | i: 1 1596 | i: 1 1597 | } 1598 | } 1599 | } 1600 | attr { 1601 | key: "padding" 1602 | value { 1603 | s: "SAME" 1604 | } 1605 | } 1606 | attr { 1607 | key: "strides" 1608 | value { 1609 | list { 1610 | i: 1 1611 | i: 1 1612 | i: 1 1613 | i: 1 1614 | } 1615 | } 1616 | } 1617 | } 1618 | node { 1619 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/add_1" 1620 | op: "FusedBatchNorm" 1621 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise" 1622 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/gamma" 1623 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/beta" 1624 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/moving_mean" 1625 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/moving_variance" 1626 | attr { 1627 | key: "epsilon" 1628 | value { 1629 | f: 0.0010000000475 1630 | } 1631 | } 1632 | } 1633 | node { 1634 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/Relu6" 1635 | op: "Relu6" 1636 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/add_1" 1637 | } 1638 | node { 1639 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/Conv2D" 1640 | op: "Conv2D" 1641 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/Relu6" 1642 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/weights" 1643 | attr { 1644 | key: "dilations" 1645 | value { 1646 | list { 1647 | i: 1 1648 | i: 1 1649 | i: 1 1650 | i: 1 1651 | } 1652 | } 1653 | } 1654 | attr { 1655 | key: "padding" 1656 | value { 1657 | s: "SAME" 1658 | } 1659 | } 1660 | attr { 1661 | key: "strides" 1662 | value { 1663 | list { 1664 | i: 1 1665 | i: 1 1666 | i: 1 1667 | i: 1 1668 | } 1669 | } 1670 | } 1671 | } 1672 | node { 1673 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/batchnorm/add_1" 1674 | op: "FusedBatchNorm" 1675 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/Conv2D" 1676 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/gamma" 1677 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/beta" 1678 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/moving_mean" 1679 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/moving_variance" 1680 | attr { 1681 | key: "epsilon" 1682 | value { 1683 | f: 0.0010000000475 1684 | } 1685 | } 1686 | } 1687 | node { 1688 | name: "FeatureExtractor/MobilenetV2/expanded_conv_9/add" 1689 | op: "Add" 1690 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/batchnorm/add_1" 1691 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/input" 1692 | } 1693 | node { 1694 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/input" 1695 | op: "Identity" 1696 | input: "FeatureExtractor/MobilenetV2/expanded_conv_9/add" 1697 | } 1698 | node { 1699 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Conv2D" 1700 | op: "Conv2D" 1701 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/input" 1702 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/weights" 1703 | attr { 1704 | key: "dilations" 1705 | value { 1706 | list { 1707 | i: 1 1708 | i: 1 1709 | i: 1 1710 | i: 1 1711 | } 1712 | } 1713 | } 1714 | attr { 1715 | key: "padding" 1716 | value { 1717 | s: "SAME" 1718 | } 1719 | } 1720 | attr { 1721 | key: "strides" 1722 | value { 1723 | list { 1724 | i: 1 1725 | i: 1 1726 | i: 1 1727 | i: 1 1728 | } 1729 | } 1730 | } 1731 | } 1732 | node { 1733 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/add_1" 1734 | op: "FusedBatchNorm" 1735 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Conv2D" 1736 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/gamma" 1737 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/beta" 1738 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/moving_mean" 1739 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/moving_variance" 1740 | attr { 1741 | key: "epsilon" 1742 | value { 1743 | f: 0.0010000000475 1744 | } 1745 | } 1746 | } 1747 | node { 1748 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Relu6" 1749 | op: "Relu6" 1750 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/add_1" 1751 | } 1752 | node { 1753 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise" 1754 | op: "DepthwiseConv2dNative" 1755 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Relu6" 1756 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise_weights" 1757 | attr { 1758 | key: "dilations" 1759 | value { 1760 | list { 1761 | i: 1 1762 | i: 1 1763 | i: 1 1764 | i: 1 1765 | } 1766 | } 1767 | } 1768 | attr { 1769 | key: "padding" 1770 | value { 1771 | s: "SAME" 1772 | } 1773 | } 1774 | attr { 1775 | key: "strides" 1776 | value { 1777 | list { 1778 | i: 1 1779 | i: 1 1780 | i: 1 1781 | i: 1 1782 | } 1783 | } 1784 | } 1785 | } 1786 | node { 1787 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/add_1" 1788 | op: "FusedBatchNorm" 1789 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise" 1790 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/gamma" 1791 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/beta" 1792 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/moving_mean" 1793 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/moving_variance" 1794 | attr { 1795 | key: "epsilon" 1796 | value { 1797 | f: 0.0010000000475 1798 | } 1799 | } 1800 | } 1801 | node { 1802 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/Relu6" 1803 | op: "Relu6" 1804 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/add_1" 1805 | } 1806 | node { 1807 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/Conv2D" 1808 | op: "Conv2D" 1809 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/Relu6" 1810 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/weights" 1811 | attr { 1812 | key: "dilations" 1813 | value { 1814 | list { 1815 | i: 1 1816 | i: 1 1817 | i: 1 1818 | i: 1 1819 | } 1820 | } 1821 | } 1822 | attr { 1823 | key: "padding" 1824 | value { 1825 | s: "SAME" 1826 | } 1827 | } 1828 | attr { 1829 | key: "strides" 1830 | value { 1831 | list { 1832 | i: 1 1833 | i: 1 1834 | i: 1 1835 | i: 1 1836 | } 1837 | } 1838 | } 1839 | } 1840 | node { 1841 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/add_1" 1842 | op: "FusedBatchNorm" 1843 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/Conv2D" 1844 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/gamma" 1845 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/beta" 1846 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/moving_mean" 1847 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/moving_variance" 1848 | attr { 1849 | key: "epsilon" 1850 | value { 1851 | f: 0.0010000000475 1852 | } 1853 | } 1854 | } 1855 | node { 1856 | name: "FeatureExtractor/MobilenetV2/expanded_conv_10/output" 1857 | op: "Identity" 1858 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/add_1" 1859 | } 1860 | node { 1861 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Conv2D" 1862 | op: "Conv2D" 1863 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/output" 1864 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/weights" 1865 | attr { 1866 | key: "dilations" 1867 | value { 1868 | list { 1869 | i: 1 1870 | i: 1 1871 | i: 1 1872 | i: 1 1873 | } 1874 | } 1875 | } 1876 | attr { 1877 | key: "padding" 1878 | value { 1879 | s: "SAME" 1880 | } 1881 | } 1882 | attr { 1883 | key: "strides" 1884 | value { 1885 | list { 1886 | i: 1 1887 | i: 1 1888 | i: 1 1889 | i: 1 1890 | } 1891 | } 1892 | } 1893 | } 1894 | node { 1895 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/add_1" 1896 | op: "FusedBatchNorm" 1897 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Conv2D" 1898 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/gamma" 1899 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/beta" 1900 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/moving_mean" 1901 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/moving_variance" 1902 | attr { 1903 | key: "epsilon" 1904 | value { 1905 | f: 0.0010000000475 1906 | } 1907 | } 1908 | } 1909 | node { 1910 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Relu6" 1911 | op: "Relu6" 1912 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/add_1" 1913 | } 1914 | node { 1915 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise" 1916 | op: "DepthwiseConv2dNative" 1917 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Relu6" 1918 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise_weights" 1919 | attr { 1920 | key: "dilations" 1921 | value { 1922 | list { 1923 | i: 1 1924 | i: 1 1925 | i: 1 1926 | i: 1 1927 | } 1928 | } 1929 | } 1930 | attr { 1931 | key: "padding" 1932 | value { 1933 | s: "SAME" 1934 | } 1935 | } 1936 | attr { 1937 | key: "strides" 1938 | value { 1939 | list { 1940 | i: 1 1941 | i: 1 1942 | i: 1 1943 | i: 1 1944 | } 1945 | } 1946 | } 1947 | } 1948 | node { 1949 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/add_1" 1950 | op: "FusedBatchNorm" 1951 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise" 1952 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/gamma" 1953 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/beta" 1954 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/moving_mean" 1955 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/moving_variance" 1956 | attr { 1957 | key: "epsilon" 1958 | value { 1959 | f: 0.0010000000475 1960 | } 1961 | } 1962 | } 1963 | node { 1964 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/Relu6" 1965 | op: "Relu6" 1966 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/add_1" 1967 | } 1968 | node { 1969 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/Conv2D" 1970 | op: "Conv2D" 1971 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/Relu6" 1972 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/weights" 1973 | attr { 1974 | key: "dilations" 1975 | value { 1976 | list { 1977 | i: 1 1978 | i: 1 1979 | i: 1 1980 | i: 1 1981 | } 1982 | } 1983 | } 1984 | attr { 1985 | key: "padding" 1986 | value { 1987 | s: "SAME" 1988 | } 1989 | } 1990 | attr { 1991 | key: "strides" 1992 | value { 1993 | list { 1994 | i: 1 1995 | i: 1 1996 | i: 1 1997 | i: 1 1998 | } 1999 | } 2000 | } 2001 | } 2002 | node { 2003 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/batchnorm/add_1" 2004 | op: "FusedBatchNorm" 2005 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/Conv2D" 2006 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/gamma" 2007 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/beta" 2008 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/moving_mean" 2009 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/moving_variance" 2010 | attr { 2011 | key: "epsilon" 2012 | value { 2013 | f: 0.0010000000475 2014 | } 2015 | } 2016 | } 2017 | node { 2018 | name: "FeatureExtractor/MobilenetV2/expanded_conv_11/add" 2019 | op: "Add" 2020 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/batchnorm/add_1" 2021 | input: "FeatureExtractor/MobilenetV2/expanded_conv_10/output" 2022 | } 2023 | node { 2024 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/input" 2025 | op: "Identity" 2026 | input: "FeatureExtractor/MobilenetV2/expanded_conv_11/add" 2027 | } 2028 | node { 2029 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Conv2D" 2030 | op: "Conv2D" 2031 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/input" 2032 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/weights" 2033 | attr { 2034 | key: "dilations" 2035 | value { 2036 | list { 2037 | i: 1 2038 | i: 1 2039 | i: 1 2040 | i: 1 2041 | } 2042 | } 2043 | } 2044 | attr { 2045 | key: "padding" 2046 | value { 2047 | s: "SAME" 2048 | } 2049 | } 2050 | attr { 2051 | key: "strides" 2052 | value { 2053 | list { 2054 | i: 1 2055 | i: 1 2056 | i: 1 2057 | i: 1 2058 | } 2059 | } 2060 | } 2061 | } 2062 | node { 2063 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/add_1" 2064 | op: "FusedBatchNorm" 2065 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Conv2D" 2066 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/gamma" 2067 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/beta" 2068 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/moving_mean" 2069 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/moving_variance" 2070 | attr { 2071 | key: "epsilon" 2072 | value { 2073 | f: 0.0010000000475 2074 | } 2075 | } 2076 | } 2077 | node { 2078 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Relu6" 2079 | op: "Relu6" 2080 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/add_1" 2081 | } 2082 | node { 2083 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise" 2084 | op: "DepthwiseConv2dNative" 2085 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Relu6" 2086 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise_weights" 2087 | attr { 2088 | key: "dilations" 2089 | value { 2090 | list { 2091 | i: 1 2092 | i: 1 2093 | i: 1 2094 | i: 1 2095 | } 2096 | } 2097 | } 2098 | attr { 2099 | key: "padding" 2100 | value { 2101 | s: "SAME" 2102 | } 2103 | } 2104 | attr { 2105 | key: "strides" 2106 | value { 2107 | list { 2108 | i: 1 2109 | i: 1 2110 | i: 1 2111 | i: 1 2112 | } 2113 | } 2114 | } 2115 | } 2116 | node { 2117 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/add_1" 2118 | op: "FusedBatchNorm" 2119 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise" 2120 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/gamma" 2121 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/beta" 2122 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/moving_mean" 2123 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/moving_variance" 2124 | attr { 2125 | key: "epsilon" 2126 | value { 2127 | f: 0.0010000000475 2128 | } 2129 | } 2130 | } 2131 | node { 2132 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/Relu6" 2133 | op: "Relu6" 2134 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/add_1" 2135 | } 2136 | node { 2137 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/Conv2D" 2138 | op: "Conv2D" 2139 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/Relu6" 2140 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/weights" 2141 | attr { 2142 | key: "dilations" 2143 | value { 2144 | list { 2145 | i: 1 2146 | i: 1 2147 | i: 1 2148 | i: 1 2149 | } 2150 | } 2151 | } 2152 | attr { 2153 | key: "padding" 2154 | value { 2155 | s: "SAME" 2156 | } 2157 | } 2158 | attr { 2159 | key: "strides" 2160 | value { 2161 | list { 2162 | i: 1 2163 | i: 1 2164 | i: 1 2165 | i: 1 2166 | } 2167 | } 2168 | } 2169 | } 2170 | node { 2171 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/batchnorm/add_1" 2172 | op: "FusedBatchNorm" 2173 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/Conv2D" 2174 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/gamma" 2175 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/beta" 2176 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/moving_mean" 2177 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/moving_variance" 2178 | attr { 2179 | key: "epsilon" 2180 | value { 2181 | f: 0.0010000000475 2182 | } 2183 | } 2184 | } 2185 | node { 2186 | name: "FeatureExtractor/MobilenetV2/expanded_conv_12/add" 2187 | op: "Add" 2188 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/batchnorm/add_1" 2189 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/input" 2190 | } 2191 | node { 2192 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/input" 2193 | op: "Identity" 2194 | input: "FeatureExtractor/MobilenetV2/expanded_conv_12/add" 2195 | } 2196 | node { 2197 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Conv2D" 2198 | op: "Conv2D" 2199 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/input" 2200 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/weights" 2201 | attr { 2202 | key: "dilations" 2203 | value { 2204 | list { 2205 | i: 1 2206 | i: 1 2207 | i: 1 2208 | i: 1 2209 | } 2210 | } 2211 | } 2212 | attr { 2213 | key: "padding" 2214 | value { 2215 | s: "SAME" 2216 | } 2217 | } 2218 | attr { 2219 | key: "strides" 2220 | value { 2221 | list { 2222 | i: 1 2223 | i: 1 2224 | i: 1 2225 | i: 1 2226 | } 2227 | } 2228 | } 2229 | } 2230 | node { 2231 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/add_1" 2232 | op: "FusedBatchNorm" 2233 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Conv2D" 2234 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/gamma" 2235 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/beta" 2236 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/moving_mean" 2237 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/moving_variance" 2238 | attr { 2239 | key: "epsilon" 2240 | value { 2241 | f: 0.0010000000475 2242 | } 2243 | } 2244 | } 2245 | node { 2246 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6" 2247 | op: "Relu6" 2248 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/add_1" 2249 | } 2250 | node { 2251 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise" 2252 | op: "DepthwiseConv2dNative" 2253 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6" 2254 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise_weights" 2255 | attr { 2256 | key: "dilations" 2257 | value { 2258 | list { 2259 | i: 1 2260 | i: 1 2261 | i: 1 2262 | i: 1 2263 | } 2264 | } 2265 | } 2266 | attr { 2267 | key: "padding" 2268 | value { 2269 | s: "SAME" 2270 | } 2271 | } 2272 | attr { 2273 | key: "strides" 2274 | value { 2275 | list { 2276 | i: 1 2277 | i: 2 2278 | i: 2 2279 | i: 1 2280 | } 2281 | } 2282 | } 2283 | } 2284 | node { 2285 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/add_1" 2286 | op: "FusedBatchNorm" 2287 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise" 2288 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/gamma" 2289 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/beta" 2290 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/moving_mean" 2291 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/moving_variance" 2292 | attr { 2293 | key: "epsilon" 2294 | value { 2295 | f: 0.0010000000475 2296 | } 2297 | } 2298 | } 2299 | node { 2300 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/Relu6" 2301 | op: "Relu6" 2302 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/add_1" 2303 | } 2304 | node { 2305 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/Conv2D" 2306 | op: "Conv2D" 2307 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/Relu6" 2308 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/weights" 2309 | attr { 2310 | key: "dilations" 2311 | value { 2312 | list { 2313 | i: 1 2314 | i: 1 2315 | i: 1 2316 | i: 1 2317 | } 2318 | } 2319 | } 2320 | attr { 2321 | key: "padding" 2322 | value { 2323 | s: "SAME" 2324 | } 2325 | } 2326 | attr { 2327 | key: "strides" 2328 | value { 2329 | list { 2330 | i: 1 2331 | i: 1 2332 | i: 1 2333 | i: 1 2334 | } 2335 | } 2336 | } 2337 | } 2338 | node { 2339 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/add_1" 2340 | op: "FusedBatchNorm" 2341 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/Conv2D" 2342 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/gamma" 2343 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/beta" 2344 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/moving_mean" 2345 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/moving_variance" 2346 | attr { 2347 | key: "epsilon" 2348 | value { 2349 | f: 0.0010000000475 2350 | } 2351 | } 2352 | } 2353 | node { 2354 | name: "FeatureExtractor/MobilenetV2/expanded_conv_13/output" 2355 | op: "Identity" 2356 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/add_1" 2357 | } 2358 | node { 2359 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Conv2D" 2360 | op: "Conv2D" 2361 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/output" 2362 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/weights" 2363 | attr { 2364 | key: "dilations" 2365 | value { 2366 | list { 2367 | i: 1 2368 | i: 1 2369 | i: 1 2370 | i: 1 2371 | } 2372 | } 2373 | } 2374 | attr { 2375 | key: "padding" 2376 | value { 2377 | s: "SAME" 2378 | } 2379 | } 2380 | attr { 2381 | key: "strides" 2382 | value { 2383 | list { 2384 | i: 1 2385 | i: 1 2386 | i: 1 2387 | i: 1 2388 | } 2389 | } 2390 | } 2391 | } 2392 | node { 2393 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/add_1" 2394 | op: "FusedBatchNorm" 2395 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Conv2D" 2396 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/gamma" 2397 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/beta" 2398 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/moving_mean" 2399 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/moving_variance" 2400 | attr { 2401 | key: "epsilon" 2402 | value { 2403 | f: 0.0010000000475 2404 | } 2405 | } 2406 | } 2407 | node { 2408 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Relu6" 2409 | op: "Relu6" 2410 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/add_1" 2411 | } 2412 | node { 2413 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise" 2414 | op: "DepthwiseConv2dNative" 2415 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Relu6" 2416 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise_weights" 2417 | attr { 2418 | key: "dilations" 2419 | value { 2420 | list { 2421 | i: 1 2422 | i: 1 2423 | i: 1 2424 | i: 1 2425 | } 2426 | } 2427 | } 2428 | attr { 2429 | key: "padding" 2430 | value { 2431 | s: "SAME" 2432 | } 2433 | } 2434 | attr { 2435 | key: "strides" 2436 | value { 2437 | list { 2438 | i: 1 2439 | i: 1 2440 | i: 1 2441 | i: 1 2442 | } 2443 | } 2444 | } 2445 | } 2446 | node { 2447 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/add_1" 2448 | op: "FusedBatchNorm" 2449 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise" 2450 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/gamma" 2451 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/beta" 2452 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/moving_mean" 2453 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/moving_variance" 2454 | attr { 2455 | key: "epsilon" 2456 | value { 2457 | f: 0.0010000000475 2458 | } 2459 | } 2460 | } 2461 | node { 2462 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/Relu6" 2463 | op: "Relu6" 2464 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/add_1" 2465 | } 2466 | node { 2467 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/Conv2D" 2468 | op: "Conv2D" 2469 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/Relu6" 2470 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/weights" 2471 | attr { 2472 | key: "dilations" 2473 | value { 2474 | list { 2475 | i: 1 2476 | i: 1 2477 | i: 1 2478 | i: 1 2479 | } 2480 | } 2481 | } 2482 | attr { 2483 | key: "padding" 2484 | value { 2485 | s: "SAME" 2486 | } 2487 | } 2488 | attr { 2489 | key: "strides" 2490 | value { 2491 | list { 2492 | i: 1 2493 | i: 1 2494 | i: 1 2495 | i: 1 2496 | } 2497 | } 2498 | } 2499 | } 2500 | node { 2501 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/batchnorm/add_1" 2502 | op: "FusedBatchNorm" 2503 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/Conv2D" 2504 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/gamma" 2505 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/beta" 2506 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/moving_mean" 2507 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/moving_variance" 2508 | attr { 2509 | key: "epsilon" 2510 | value { 2511 | f: 0.0010000000475 2512 | } 2513 | } 2514 | } 2515 | node { 2516 | name: "FeatureExtractor/MobilenetV2/expanded_conv_14/add" 2517 | op: "Add" 2518 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/batchnorm/add_1" 2519 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/output" 2520 | } 2521 | node { 2522 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/input" 2523 | op: "Identity" 2524 | input: "FeatureExtractor/MobilenetV2/expanded_conv_14/add" 2525 | } 2526 | node { 2527 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Conv2D" 2528 | op: "Conv2D" 2529 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/input" 2530 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/weights" 2531 | attr { 2532 | key: "dilations" 2533 | value { 2534 | list { 2535 | i: 1 2536 | i: 1 2537 | i: 1 2538 | i: 1 2539 | } 2540 | } 2541 | } 2542 | attr { 2543 | key: "padding" 2544 | value { 2545 | s: "SAME" 2546 | } 2547 | } 2548 | attr { 2549 | key: "strides" 2550 | value { 2551 | list { 2552 | i: 1 2553 | i: 1 2554 | i: 1 2555 | i: 1 2556 | } 2557 | } 2558 | } 2559 | } 2560 | node { 2561 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/add_1" 2562 | op: "FusedBatchNorm" 2563 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Conv2D" 2564 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/gamma" 2565 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/beta" 2566 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/moving_mean" 2567 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/moving_variance" 2568 | attr { 2569 | key: "epsilon" 2570 | value { 2571 | f: 0.0010000000475 2572 | } 2573 | } 2574 | } 2575 | node { 2576 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Relu6" 2577 | op: "Relu6" 2578 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/add_1" 2579 | } 2580 | node { 2581 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise" 2582 | op: "DepthwiseConv2dNative" 2583 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Relu6" 2584 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise_weights" 2585 | attr { 2586 | key: "dilations" 2587 | value { 2588 | list { 2589 | i: 1 2590 | i: 1 2591 | i: 1 2592 | i: 1 2593 | } 2594 | } 2595 | } 2596 | attr { 2597 | key: "padding" 2598 | value { 2599 | s: "SAME" 2600 | } 2601 | } 2602 | attr { 2603 | key: "strides" 2604 | value { 2605 | list { 2606 | i: 1 2607 | i: 1 2608 | i: 1 2609 | i: 1 2610 | } 2611 | } 2612 | } 2613 | } 2614 | node { 2615 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/add_1" 2616 | op: "FusedBatchNorm" 2617 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise" 2618 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/gamma" 2619 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/beta" 2620 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/moving_mean" 2621 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/moving_variance" 2622 | attr { 2623 | key: "epsilon" 2624 | value { 2625 | f: 0.0010000000475 2626 | } 2627 | } 2628 | } 2629 | node { 2630 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/Relu6" 2631 | op: "Relu6" 2632 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/add_1" 2633 | } 2634 | node { 2635 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/Conv2D" 2636 | op: "Conv2D" 2637 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/Relu6" 2638 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/weights" 2639 | attr { 2640 | key: "dilations" 2641 | value { 2642 | list { 2643 | i: 1 2644 | i: 1 2645 | i: 1 2646 | i: 1 2647 | } 2648 | } 2649 | } 2650 | attr { 2651 | key: "padding" 2652 | value { 2653 | s: "SAME" 2654 | } 2655 | } 2656 | attr { 2657 | key: "strides" 2658 | value { 2659 | list { 2660 | i: 1 2661 | i: 1 2662 | i: 1 2663 | i: 1 2664 | } 2665 | } 2666 | } 2667 | } 2668 | node { 2669 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/batchnorm/add_1" 2670 | op: "FusedBatchNorm" 2671 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/Conv2D" 2672 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/gamma" 2673 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/beta" 2674 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/moving_mean" 2675 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/moving_variance" 2676 | attr { 2677 | key: "epsilon" 2678 | value { 2679 | f: 0.0010000000475 2680 | } 2681 | } 2682 | } 2683 | node { 2684 | name: "FeatureExtractor/MobilenetV2/expanded_conv_15/add" 2685 | op: "Add" 2686 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/batchnorm/add_1" 2687 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/input" 2688 | } 2689 | node { 2690 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/input" 2691 | op: "Identity" 2692 | input: "FeatureExtractor/MobilenetV2/expanded_conv_15/add" 2693 | } 2694 | node { 2695 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Conv2D" 2696 | op: "Conv2D" 2697 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/input" 2698 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/weights" 2699 | attr { 2700 | key: "dilations" 2701 | value { 2702 | list { 2703 | i: 1 2704 | i: 1 2705 | i: 1 2706 | i: 1 2707 | } 2708 | } 2709 | } 2710 | attr { 2711 | key: "padding" 2712 | value { 2713 | s: "SAME" 2714 | } 2715 | } 2716 | attr { 2717 | key: "strides" 2718 | value { 2719 | list { 2720 | i: 1 2721 | i: 1 2722 | i: 1 2723 | i: 1 2724 | } 2725 | } 2726 | } 2727 | } 2728 | node { 2729 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/add_1" 2730 | op: "FusedBatchNorm" 2731 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Conv2D" 2732 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/gamma" 2733 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/beta" 2734 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/moving_mean" 2735 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/moving_variance" 2736 | attr { 2737 | key: "epsilon" 2738 | value { 2739 | f: 0.0010000000475 2740 | } 2741 | } 2742 | } 2743 | node { 2744 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Relu6" 2745 | op: "Relu6" 2746 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/add_1" 2747 | } 2748 | node { 2749 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise" 2750 | op: "DepthwiseConv2dNative" 2751 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Relu6" 2752 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise_weights" 2753 | attr { 2754 | key: "dilations" 2755 | value { 2756 | list { 2757 | i: 1 2758 | i: 1 2759 | i: 1 2760 | i: 1 2761 | } 2762 | } 2763 | } 2764 | attr { 2765 | key: "padding" 2766 | value { 2767 | s: "SAME" 2768 | } 2769 | } 2770 | attr { 2771 | key: "strides" 2772 | value { 2773 | list { 2774 | i: 1 2775 | i: 1 2776 | i: 1 2777 | i: 1 2778 | } 2779 | } 2780 | } 2781 | } 2782 | node { 2783 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/add_1" 2784 | op: "FusedBatchNorm" 2785 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise" 2786 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/gamma" 2787 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/beta" 2788 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/moving_mean" 2789 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/moving_variance" 2790 | attr { 2791 | key: "epsilon" 2792 | value { 2793 | f: 0.0010000000475 2794 | } 2795 | } 2796 | } 2797 | node { 2798 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/Relu6" 2799 | op: "Relu6" 2800 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/add_1" 2801 | } 2802 | node { 2803 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/Conv2D" 2804 | op: "Conv2D" 2805 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/Relu6" 2806 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/weights" 2807 | attr { 2808 | key: "dilations" 2809 | value { 2810 | list { 2811 | i: 1 2812 | i: 1 2813 | i: 1 2814 | i: 1 2815 | } 2816 | } 2817 | } 2818 | attr { 2819 | key: "padding" 2820 | value { 2821 | s: "SAME" 2822 | } 2823 | } 2824 | attr { 2825 | key: "strides" 2826 | value { 2827 | list { 2828 | i: 1 2829 | i: 1 2830 | i: 1 2831 | i: 1 2832 | } 2833 | } 2834 | } 2835 | } 2836 | node { 2837 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/add_1" 2838 | op: "FusedBatchNorm" 2839 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/Conv2D" 2840 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/gamma" 2841 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/beta" 2842 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/moving_mean" 2843 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/moving_variance" 2844 | attr { 2845 | key: "epsilon" 2846 | value { 2847 | f: 0.0010000000475 2848 | } 2849 | } 2850 | } 2851 | node { 2852 | name: "FeatureExtractor/MobilenetV2/expanded_conv_16/output" 2853 | op: "Identity" 2854 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/add_1" 2855 | } 2856 | node { 2857 | name: "FeatureExtractor/MobilenetV2/Conv_1/Conv2D" 2858 | op: "Conv2D" 2859 | input: "FeatureExtractor/MobilenetV2/expanded_conv_16/output" 2860 | input: "FeatureExtractor/MobilenetV2/Conv_1/weights" 2861 | attr { 2862 | key: "dilations" 2863 | value { 2864 | list { 2865 | i: 1 2866 | i: 1 2867 | i: 1 2868 | i: 1 2869 | } 2870 | } 2871 | } 2872 | attr { 2873 | key: "padding" 2874 | value { 2875 | s: "SAME" 2876 | } 2877 | } 2878 | attr { 2879 | key: "strides" 2880 | value { 2881 | list { 2882 | i: 1 2883 | i: 1 2884 | i: 1 2885 | i: 1 2886 | } 2887 | } 2888 | } 2889 | } 2890 | node { 2891 | name: "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/add_1" 2892 | op: "FusedBatchNorm" 2893 | input: "FeatureExtractor/MobilenetV2/Conv_1/Conv2D" 2894 | input: "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/gamma" 2895 | input: "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/beta" 2896 | input: "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/moving_mean" 2897 | input: "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/moving_variance" 2898 | attr { 2899 | key: "epsilon" 2900 | value { 2901 | f: 0.0010000000475 2902 | } 2903 | } 2904 | } 2905 | node { 2906 | name: "FeatureExtractor/MobilenetV2/Conv_1/Relu6" 2907 | op: "Relu6" 2908 | input: "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/add_1" 2909 | } 2910 | node { 2911 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Conv2D" 2912 | op: "Conv2D" 2913 | input: "FeatureExtractor/MobilenetV2/Conv_1/Relu6" 2914 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/weights" 2915 | attr { 2916 | key: "dilations" 2917 | value { 2918 | list { 2919 | i: 1 2920 | i: 1 2921 | i: 1 2922 | i: 1 2923 | } 2924 | } 2925 | } 2926 | attr { 2927 | key: "padding" 2928 | value { 2929 | s: "SAME" 2930 | } 2931 | } 2932 | attr { 2933 | key: "strides" 2934 | value { 2935 | list { 2936 | i: 1 2937 | i: 1 2938 | i: 1 2939 | i: 1 2940 | } 2941 | } 2942 | } 2943 | } 2944 | node { 2945 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/add_1" 2946 | op: "FusedBatchNorm" 2947 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Conv2D" 2948 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/gamma" 2949 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/beta" 2950 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/moving_mean" 2951 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/moving_variance" 2952 | attr { 2953 | key: "epsilon" 2954 | value { 2955 | f: 0.0010000000475 2956 | } 2957 | } 2958 | } 2959 | node { 2960 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Relu6" 2961 | op: "Relu6" 2962 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/add_1" 2963 | } 2964 | node { 2965 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise" 2966 | op: "DepthwiseConv2dNative" 2967 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Relu6" 2968 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise_weights" 2969 | attr { 2970 | key: "dilations" 2971 | value { 2972 | list { 2973 | i: 1 2974 | i: 1 2975 | i: 1 2976 | i: 1 2977 | } 2978 | } 2979 | } 2980 | attr { 2981 | key: "padding" 2982 | value { 2983 | s: "SAME" 2984 | } 2985 | } 2986 | attr { 2987 | key: "strides" 2988 | value { 2989 | list { 2990 | i: 1 2991 | i: 2 2992 | i: 2 2993 | i: 1 2994 | } 2995 | } 2996 | } 2997 | } 2998 | node { 2999 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/add_1" 3000 | op: "FusedBatchNorm" 3001 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise" 3002 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/gamma" 3003 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/beta" 3004 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/moving_mean" 3005 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/moving_variance" 3006 | attr { 3007 | key: "epsilon" 3008 | value { 3009 | f: 0.0010000000475 3010 | } 3011 | } 3012 | } 3013 | node { 3014 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/Relu6" 3015 | op: "Relu6" 3016 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/add_1" 3017 | } 3018 | node { 3019 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Conv2D" 3020 | op: "Conv2D" 3021 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/Relu6" 3022 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/weights" 3023 | attr { 3024 | key: "dilations" 3025 | value { 3026 | list { 3027 | i: 1 3028 | i: 1 3029 | i: 1 3030 | i: 1 3031 | } 3032 | } 3033 | } 3034 | attr { 3035 | key: "padding" 3036 | value { 3037 | s: "SAME" 3038 | } 3039 | } 3040 | attr { 3041 | key: "strides" 3042 | value { 3043 | list { 3044 | i: 1 3045 | i: 1 3046 | i: 1 3047 | i: 1 3048 | } 3049 | } 3050 | } 3051 | } 3052 | node { 3053 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/add_1" 3054 | op: "FusedBatchNorm" 3055 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Conv2D" 3056 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/gamma" 3057 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/beta" 3058 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/moving_mean" 3059 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/moving_variance" 3060 | attr { 3061 | key: "epsilon" 3062 | value { 3063 | f: 0.0010000000475 3064 | } 3065 | } 3066 | } 3067 | node { 3068 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6" 3069 | op: "Relu6" 3070 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/add_1" 3071 | } 3072 | node { 3073 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Conv2D" 3074 | op: "Conv2D" 3075 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6" 3076 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/weights" 3077 | attr { 3078 | key: "dilations" 3079 | value { 3080 | list { 3081 | i: 1 3082 | i: 1 3083 | i: 1 3084 | i: 1 3085 | } 3086 | } 3087 | } 3088 | attr { 3089 | key: "padding" 3090 | value { 3091 | s: "SAME" 3092 | } 3093 | } 3094 | attr { 3095 | key: "strides" 3096 | value { 3097 | list { 3098 | i: 1 3099 | i: 1 3100 | i: 1 3101 | i: 1 3102 | } 3103 | } 3104 | } 3105 | } 3106 | node { 3107 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/add_1" 3108 | op: "FusedBatchNorm" 3109 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Conv2D" 3110 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/gamma" 3111 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/beta" 3112 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/moving_mean" 3113 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/moving_variance" 3114 | attr { 3115 | key: "epsilon" 3116 | value { 3117 | f: 0.0010000000475 3118 | } 3119 | } 3120 | } 3121 | node { 3122 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Relu6" 3123 | op: "Relu6" 3124 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/add_1" 3125 | } 3126 | node { 3127 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise" 3128 | op: "DepthwiseConv2dNative" 3129 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Relu6" 3130 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise_weights" 3131 | attr { 3132 | key: "dilations" 3133 | value { 3134 | list { 3135 | i: 1 3136 | i: 1 3137 | i: 1 3138 | i: 1 3139 | } 3140 | } 3141 | } 3142 | attr { 3143 | key: "padding" 3144 | value { 3145 | s: "SAME" 3146 | } 3147 | } 3148 | attr { 3149 | key: "strides" 3150 | value { 3151 | list { 3152 | i: 1 3153 | i: 2 3154 | i: 2 3155 | i: 1 3156 | } 3157 | } 3158 | } 3159 | } 3160 | node { 3161 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1" 3162 | op: "FusedBatchNorm" 3163 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise" 3164 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/gamma" 3165 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/beta" 3166 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/moving_mean" 3167 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/moving_variance" 3168 | attr { 3169 | key: "epsilon" 3170 | value { 3171 | f: 0.0010000000475 3172 | } 3173 | } 3174 | } 3175 | node { 3176 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/Relu6" 3177 | op: "Relu6" 3178 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1" 3179 | } 3180 | node { 3181 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Conv2D" 3182 | op: "Conv2D" 3183 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/Relu6" 3184 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/weights" 3185 | attr { 3186 | key: "dilations" 3187 | value { 3188 | list { 3189 | i: 1 3190 | i: 1 3191 | i: 1 3192 | i: 1 3193 | } 3194 | } 3195 | } 3196 | attr { 3197 | key: "padding" 3198 | value { 3199 | s: "SAME" 3200 | } 3201 | } 3202 | attr { 3203 | key: "strides" 3204 | value { 3205 | list { 3206 | i: 1 3207 | i: 1 3208 | i: 1 3209 | i: 1 3210 | } 3211 | } 3212 | } 3213 | } 3214 | node { 3215 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/add_1" 3216 | op: "FusedBatchNorm" 3217 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Conv2D" 3218 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/gamma" 3219 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/beta" 3220 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/moving_mean" 3221 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/moving_variance" 3222 | attr { 3223 | key: "epsilon" 3224 | value { 3225 | f: 0.0010000000475 3226 | } 3227 | } 3228 | } 3229 | node { 3230 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6" 3231 | op: "Relu6" 3232 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/add_1" 3233 | } 3234 | node { 3235 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Conv2D" 3236 | op: "Conv2D" 3237 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6" 3238 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/weights" 3239 | attr { 3240 | key: "dilations" 3241 | value { 3242 | list { 3243 | i: 1 3244 | i: 1 3245 | i: 1 3246 | i: 1 3247 | } 3248 | } 3249 | } 3250 | attr { 3251 | key: "padding" 3252 | value { 3253 | s: "SAME" 3254 | } 3255 | } 3256 | attr { 3257 | key: "strides" 3258 | value { 3259 | list { 3260 | i: 1 3261 | i: 1 3262 | i: 1 3263 | i: 1 3264 | } 3265 | } 3266 | } 3267 | } 3268 | node { 3269 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/add_1" 3270 | op: "FusedBatchNorm" 3271 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Conv2D" 3272 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/gamma" 3273 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/beta" 3274 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/moving_mean" 3275 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/moving_variance" 3276 | attr { 3277 | key: "epsilon" 3278 | value { 3279 | f: 0.0010000000475 3280 | } 3281 | } 3282 | } 3283 | node { 3284 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Relu6" 3285 | op: "Relu6" 3286 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/add_1" 3287 | } 3288 | node { 3289 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise" 3290 | op: "DepthwiseConv2dNative" 3291 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Relu6" 3292 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise_weights" 3293 | attr { 3294 | key: "dilations" 3295 | value { 3296 | list { 3297 | i: 1 3298 | i: 1 3299 | i: 1 3300 | i: 1 3301 | } 3302 | } 3303 | } 3304 | attr { 3305 | key: "padding" 3306 | value { 3307 | s: "SAME" 3308 | } 3309 | } 3310 | attr { 3311 | key: "strides" 3312 | value { 3313 | list { 3314 | i: 1 3315 | i: 2 3316 | i: 2 3317 | i: 1 3318 | } 3319 | } 3320 | } 3321 | } 3322 | node { 3323 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1" 3324 | op: "FusedBatchNorm" 3325 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise" 3326 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/gamma" 3327 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/beta" 3328 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/moving_mean" 3329 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/moving_variance" 3330 | attr { 3331 | key: "epsilon" 3332 | value { 3333 | f: 0.0010000000475 3334 | } 3335 | } 3336 | } 3337 | node { 3338 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/Relu6" 3339 | op: "Relu6" 3340 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1" 3341 | } 3342 | node { 3343 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Conv2D" 3344 | op: "Conv2D" 3345 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/Relu6" 3346 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/weights" 3347 | attr { 3348 | key: "dilations" 3349 | value { 3350 | list { 3351 | i: 1 3352 | i: 1 3353 | i: 1 3354 | i: 1 3355 | } 3356 | } 3357 | } 3358 | attr { 3359 | key: "padding" 3360 | value { 3361 | s: "SAME" 3362 | } 3363 | } 3364 | attr { 3365 | key: "strides" 3366 | value { 3367 | list { 3368 | i: 1 3369 | i: 1 3370 | i: 1 3371 | i: 1 3372 | } 3373 | } 3374 | } 3375 | } 3376 | node { 3377 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/add_1" 3378 | op: "FusedBatchNorm" 3379 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Conv2D" 3380 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/gamma" 3381 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/beta" 3382 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/moving_mean" 3383 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/moving_variance" 3384 | attr { 3385 | key: "epsilon" 3386 | value { 3387 | f: 0.0010000000475 3388 | } 3389 | } 3390 | } 3391 | node { 3392 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6" 3393 | op: "Relu6" 3394 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/add_1" 3395 | } 3396 | node { 3397 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Conv2D" 3398 | op: "Conv2D" 3399 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6" 3400 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/weights" 3401 | attr { 3402 | key: "dilations" 3403 | value { 3404 | list { 3405 | i: 1 3406 | i: 1 3407 | i: 1 3408 | i: 1 3409 | } 3410 | } 3411 | } 3412 | attr { 3413 | key: "padding" 3414 | value { 3415 | s: "SAME" 3416 | } 3417 | } 3418 | attr { 3419 | key: "strides" 3420 | value { 3421 | list { 3422 | i: 1 3423 | i: 1 3424 | i: 1 3425 | i: 1 3426 | } 3427 | } 3428 | } 3429 | } 3430 | node { 3431 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/add_1" 3432 | op: "FusedBatchNorm" 3433 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Conv2D" 3434 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/gamma" 3435 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/beta" 3436 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/moving_mean" 3437 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/moving_variance" 3438 | attr { 3439 | key: "epsilon" 3440 | value { 3441 | f: 0.0010000000475 3442 | } 3443 | } 3444 | } 3445 | node { 3446 | name: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Relu6" 3447 | op: "Relu6" 3448 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/add_1" 3449 | } 3450 | node { 3451 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise" 3452 | op: "DepthwiseConv2dNative" 3453 | input: "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Relu6" 3454 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise_weights" 3455 | attr { 3456 | key: "dilations" 3457 | value { 3458 | list { 3459 | i: 1 3460 | i: 1 3461 | i: 1 3462 | i: 1 3463 | } 3464 | } 3465 | } 3466 | attr { 3467 | key: "padding" 3468 | value { 3469 | s: "SAME" 3470 | } 3471 | } 3472 | attr { 3473 | key: "strides" 3474 | value { 3475 | list { 3476 | i: 1 3477 | i: 2 3478 | i: 2 3479 | i: 1 3480 | } 3481 | } 3482 | } 3483 | } 3484 | node { 3485 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/add_1" 3486 | op: "FusedBatchNorm" 3487 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise" 3488 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/gamma" 3489 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/beta" 3490 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/moving_mean" 3491 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/moving_variance" 3492 | attr { 3493 | key: "epsilon" 3494 | value { 3495 | f: 0.0010000000475 3496 | } 3497 | } 3498 | } 3499 | node { 3500 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/Relu6" 3501 | op: "Relu6" 3502 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/add_1" 3503 | } 3504 | node { 3505 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Conv2D" 3506 | op: "Conv2D" 3507 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/Relu6" 3508 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/weights" 3509 | attr { 3510 | key: "dilations" 3511 | value { 3512 | list { 3513 | i: 1 3514 | i: 1 3515 | i: 1 3516 | i: 1 3517 | } 3518 | } 3519 | } 3520 | attr { 3521 | key: "padding" 3522 | value { 3523 | s: "SAME" 3524 | } 3525 | } 3526 | attr { 3527 | key: "strides" 3528 | value { 3529 | list { 3530 | i: 1 3531 | i: 1 3532 | i: 1 3533 | i: 1 3534 | } 3535 | } 3536 | } 3537 | } 3538 | node { 3539 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/add_1" 3540 | op: "FusedBatchNorm" 3541 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Conv2D" 3542 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/gamma" 3543 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/beta" 3544 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/moving_mean" 3545 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/moving_variance" 3546 | attr { 3547 | key: "epsilon" 3548 | value { 3549 | f: 0.0010000000475 3550 | } 3551 | } 3552 | } 3553 | node { 3554 | name: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Relu6" 3555 | op: "Relu6" 3556 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/add_1" 3557 | } 3558 | node { 3559 | name: "BoxPredictor_0/BoxEncodingPredictor/Conv2D" 3560 | op: "Conv2D" 3561 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6" 3562 | input: "BoxPredictor_0/BoxEncodingPredictor/weights" 3563 | attr { 3564 | key: "dilations" 3565 | value { 3566 | list { 3567 | i: 1 3568 | i: 1 3569 | i: 1 3570 | i: 1 3571 | } 3572 | } 3573 | } 3574 | attr { 3575 | key: "loc_pred_transposed" 3576 | value { 3577 | b: true 3578 | } 3579 | } 3580 | attr { 3581 | key: "padding" 3582 | value { 3583 | s: "SAME" 3584 | } 3585 | } 3586 | attr { 3587 | key: "strides" 3588 | value { 3589 | list { 3590 | i: 1 3591 | i: 1 3592 | i: 1 3593 | i: 1 3594 | } 3595 | } 3596 | } 3597 | } 3598 | node { 3599 | name: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd" 3600 | op: "BiasAdd" 3601 | input: "BoxPredictor_0/BoxEncodingPredictor/Conv2D" 3602 | input: "BoxPredictor_0/BoxEncodingPredictor/biases" 3603 | } 3604 | node { 3605 | name: "BoxPredictor_0/ClassPredictor/Conv2D" 3606 | op: "Conv2D" 3607 | input: "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6" 3608 | input: "BoxPredictor_0/ClassPredictor/weights" 3609 | attr { 3610 | key: "dilations" 3611 | value { 3612 | list { 3613 | i: 1 3614 | i: 1 3615 | i: 1 3616 | i: 1 3617 | } 3618 | } 3619 | } 3620 | attr { 3621 | key: "padding" 3622 | value { 3623 | s: "SAME" 3624 | } 3625 | } 3626 | attr { 3627 | key: "strides" 3628 | value { 3629 | list { 3630 | i: 1 3631 | i: 1 3632 | i: 1 3633 | i: 1 3634 | } 3635 | } 3636 | } 3637 | } 3638 | node { 3639 | name: "BoxPredictor_0/ClassPredictor/BiasAdd" 3640 | op: "BiasAdd" 3641 | input: "BoxPredictor_0/ClassPredictor/Conv2D" 3642 | input: "BoxPredictor_0/ClassPredictor/biases" 3643 | } 3644 | node { 3645 | name: "BoxPredictor_1/BoxEncodingPredictor/Conv2D" 3646 | op: "Conv2D" 3647 | input: "FeatureExtractor/MobilenetV2/Conv_1/Relu6" 3648 | input: "BoxPredictor_1/BoxEncodingPredictor/weights" 3649 | attr { 3650 | key: "dilations" 3651 | value { 3652 | list { 3653 | i: 1 3654 | i: 1 3655 | i: 1 3656 | i: 1 3657 | } 3658 | } 3659 | } 3660 | attr { 3661 | key: "loc_pred_transposed" 3662 | value { 3663 | b: true 3664 | } 3665 | } 3666 | attr { 3667 | key: "padding" 3668 | value { 3669 | s: "SAME" 3670 | } 3671 | } 3672 | attr { 3673 | key: "strides" 3674 | value { 3675 | list { 3676 | i: 1 3677 | i: 1 3678 | i: 1 3679 | i: 1 3680 | } 3681 | } 3682 | } 3683 | } 3684 | node { 3685 | name: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd" 3686 | op: "BiasAdd" 3687 | input: "BoxPredictor_1/BoxEncodingPredictor/Conv2D" 3688 | input: "BoxPredictor_1/BoxEncodingPredictor/biases" 3689 | } 3690 | node { 3691 | name: "BoxPredictor_1/ClassPredictor/Conv2D" 3692 | op: "Conv2D" 3693 | input: "FeatureExtractor/MobilenetV2/Conv_1/Relu6" 3694 | input: "BoxPredictor_1/ClassPredictor/weights" 3695 | attr { 3696 | key: "dilations" 3697 | value { 3698 | list { 3699 | i: 1 3700 | i: 1 3701 | i: 1 3702 | i: 1 3703 | } 3704 | } 3705 | } 3706 | attr { 3707 | key: "padding" 3708 | value { 3709 | s: "SAME" 3710 | } 3711 | } 3712 | attr { 3713 | key: "strides" 3714 | value { 3715 | list { 3716 | i: 1 3717 | i: 1 3718 | i: 1 3719 | i: 1 3720 | } 3721 | } 3722 | } 3723 | } 3724 | node { 3725 | name: "BoxPredictor_1/ClassPredictor/BiasAdd" 3726 | op: "BiasAdd" 3727 | input: "BoxPredictor_1/ClassPredictor/Conv2D" 3728 | input: "BoxPredictor_1/ClassPredictor/biases" 3729 | } 3730 | node { 3731 | name: "BoxPredictor_2/BoxEncodingPredictor/Conv2D" 3732 | op: "Conv2D" 3733 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6" 3734 | input: "BoxPredictor_2/BoxEncodingPredictor/weights" 3735 | attr { 3736 | key: "dilations" 3737 | value { 3738 | list { 3739 | i: 1 3740 | i: 1 3741 | i: 1 3742 | i: 1 3743 | } 3744 | } 3745 | } 3746 | attr { 3747 | key: "loc_pred_transposed" 3748 | value { 3749 | b: true 3750 | } 3751 | } 3752 | attr { 3753 | key: "padding" 3754 | value { 3755 | s: "SAME" 3756 | } 3757 | } 3758 | attr { 3759 | key: "strides" 3760 | value { 3761 | list { 3762 | i: 1 3763 | i: 1 3764 | i: 1 3765 | i: 1 3766 | } 3767 | } 3768 | } 3769 | } 3770 | node { 3771 | name: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd" 3772 | op: "BiasAdd" 3773 | input: "BoxPredictor_2/BoxEncodingPredictor/Conv2D" 3774 | input: "BoxPredictor_2/BoxEncodingPredictor/biases" 3775 | } 3776 | node { 3777 | name: "BoxPredictor_2/ClassPredictor/Conv2D" 3778 | op: "Conv2D" 3779 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6" 3780 | input: "BoxPredictor_2/ClassPredictor/weights" 3781 | attr { 3782 | key: "dilations" 3783 | value { 3784 | list { 3785 | i: 1 3786 | i: 1 3787 | i: 1 3788 | i: 1 3789 | } 3790 | } 3791 | } 3792 | attr { 3793 | key: "padding" 3794 | value { 3795 | s: "SAME" 3796 | } 3797 | } 3798 | attr { 3799 | key: "strides" 3800 | value { 3801 | list { 3802 | i: 1 3803 | i: 1 3804 | i: 1 3805 | i: 1 3806 | } 3807 | } 3808 | } 3809 | } 3810 | node { 3811 | name: "BoxPredictor_2/ClassPredictor/BiasAdd" 3812 | op: "BiasAdd" 3813 | input: "BoxPredictor_2/ClassPredictor/Conv2D" 3814 | input: "BoxPredictor_2/ClassPredictor/biases" 3815 | } 3816 | node { 3817 | name: "BoxPredictor_3/BoxEncodingPredictor/Conv2D" 3818 | op: "Conv2D" 3819 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6" 3820 | input: "BoxPredictor_3/BoxEncodingPredictor/weights" 3821 | attr { 3822 | key: "dilations" 3823 | value { 3824 | list { 3825 | i: 1 3826 | i: 1 3827 | i: 1 3828 | i: 1 3829 | } 3830 | } 3831 | } 3832 | attr { 3833 | key: "loc_pred_transposed" 3834 | value { 3835 | b: true 3836 | } 3837 | } 3838 | attr { 3839 | key: "padding" 3840 | value { 3841 | s: "SAME" 3842 | } 3843 | } 3844 | attr { 3845 | key: "strides" 3846 | value { 3847 | list { 3848 | i: 1 3849 | i: 1 3850 | i: 1 3851 | i: 1 3852 | } 3853 | } 3854 | } 3855 | } 3856 | node { 3857 | name: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd" 3858 | op: "BiasAdd" 3859 | input: "BoxPredictor_3/BoxEncodingPredictor/Conv2D" 3860 | input: "BoxPredictor_3/BoxEncodingPredictor/biases" 3861 | } 3862 | node { 3863 | name: "BoxPredictor_3/ClassPredictor/Conv2D" 3864 | op: "Conv2D" 3865 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6" 3866 | input: "BoxPredictor_3/ClassPredictor/weights" 3867 | attr { 3868 | key: "dilations" 3869 | value { 3870 | list { 3871 | i: 1 3872 | i: 1 3873 | i: 1 3874 | i: 1 3875 | } 3876 | } 3877 | } 3878 | attr { 3879 | key: "padding" 3880 | value { 3881 | s: "SAME" 3882 | } 3883 | } 3884 | attr { 3885 | key: "strides" 3886 | value { 3887 | list { 3888 | i: 1 3889 | i: 1 3890 | i: 1 3891 | i: 1 3892 | } 3893 | } 3894 | } 3895 | } 3896 | node { 3897 | name: "BoxPredictor_3/ClassPredictor/BiasAdd" 3898 | op: "BiasAdd" 3899 | input: "BoxPredictor_3/ClassPredictor/Conv2D" 3900 | input: "BoxPredictor_3/ClassPredictor/biases" 3901 | } 3902 | node { 3903 | name: "BoxPredictor_4/BoxEncodingPredictor/Conv2D" 3904 | op: "Conv2D" 3905 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6" 3906 | input: "BoxPredictor_4/BoxEncodingPredictor/weights" 3907 | attr { 3908 | key: "dilations" 3909 | value { 3910 | list { 3911 | i: 1 3912 | i: 1 3913 | i: 1 3914 | i: 1 3915 | } 3916 | } 3917 | } 3918 | attr { 3919 | key: "loc_pred_transposed" 3920 | value { 3921 | b: true 3922 | } 3923 | } 3924 | attr { 3925 | key: "padding" 3926 | value { 3927 | s: "SAME" 3928 | } 3929 | } 3930 | attr { 3931 | key: "strides" 3932 | value { 3933 | list { 3934 | i: 1 3935 | i: 1 3936 | i: 1 3937 | i: 1 3938 | } 3939 | } 3940 | } 3941 | } 3942 | node { 3943 | name: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd" 3944 | op: "BiasAdd" 3945 | input: "BoxPredictor_4/BoxEncodingPredictor/Conv2D" 3946 | input: "BoxPredictor_4/BoxEncodingPredictor/biases" 3947 | } 3948 | node { 3949 | name: "BoxPredictor_4/ClassPredictor/Conv2D" 3950 | op: "Conv2D" 3951 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6" 3952 | input: "BoxPredictor_4/ClassPredictor/weights" 3953 | attr { 3954 | key: "dilations" 3955 | value { 3956 | list { 3957 | i: 1 3958 | i: 1 3959 | i: 1 3960 | i: 1 3961 | } 3962 | } 3963 | } 3964 | attr { 3965 | key: "padding" 3966 | value { 3967 | s: "SAME" 3968 | } 3969 | } 3970 | attr { 3971 | key: "strides" 3972 | value { 3973 | list { 3974 | i: 1 3975 | i: 1 3976 | i: 1 3977 | i: 1 3978 | } 3979 | } 3980 | } 3981 | } 3982 | node { 3983 | name: "BoxPredictor_4/ClassPredictor/BiasAdd" 3984 | op: "BiasAdd" 3985 | input: "BoxPredictor_4/ClassPredictor/Conv2D" 3986 | input: "BoxPredictor_4/ClassPredictor/biases" 3987 | } 3988 | node { 3989 | name: "BoxPredictor_5/BoxEncodingPredictor/Conv2D" 3990 | op: "Conv2D" 3991 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Relu6" 3992 | input: "BoxPredictor_5/BoxEncodingPredictor/weights" 3993 | attr { 3994 | key: "dilations" 3995 | value { 3996 | list { 3997 | i: 1 3998 | i: 1 3999 | i: 1 4000 | i: 1 4001 | } 4002 | } 4003 | } 4004 | attr { 4005 | key: "loc_pred_transposed" 4006 | value { 4007 | b: true 4008 | } 4009 | } 4010 | attr { 4011 | key: "padding" 4012 | value { 4013 | s: "SAME" 4014 | } 4015 | } 4016 | attr { 4017 | key: "strides" 4018 | value { 4019 | list { 4020 | i: 1 4021 | i: 1 4022 | i: 1 4023 | i: 1 4024 | } 4025 | } 4026 | } 4027 | } 4028 | node { 4029 | name: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd" 4030 | op: "BiasAdd" 4031 | input: "BoxPredictor_5/BoxEncodingPredictor/Conv2D" 4032 | input: "BoxPredictor_5/BoxEncodingPredictor/biases" 4033 | } 4034 | node { 4035 | name: "BoxPredictor_5/ClassPredictor/Conv2D" 4036 | op: "Conv2D" 4037 | input: "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Relu6" 4038 | input: "BoxPredictor_5/ClassPredictor/weights" 4039 | attr { 4040 | key: "dilations" 4041 | value { 4042 | list { 4043 | i: 1 4044 | i: 1 4045 | i: 1 4046 | i: 1 4047 | } 4048 | } 4049 | } 4050 | attr { 4051 | key: "padding" 4052 | value { 4053 | s: "SAME" 4054 | } 4055 | } 4056 | attr { 4057 | key: "strides" 4058 | value { 4059 | list { 4060 | i: 1 4061 | i: 1 4062 | i: 1 4063 | i: 1 4064 | } 4065 | } 4066 | } 4067 | } 4068 | node { 4069 | name: "BoxPredictor_5/ClassPredictor/BiasAdd" 4070 | op: "BiasAdd" 4071 | input: "BoxPredictor_5/ClassPredictor/Conv2D" 4072 | input: "BoxPredictor_5/ClassPredictor/biases" 4073 | } 4074 | node { 4075 | name: "concat/axis_flatten" 4076 | op: "Const" 4077 | attr { 4078 | key: "value" 4079 | value { 4080 | tensor { 4081 | dtype: DT_INT32 4082 | tensor_shape { 4083 | dim { 4084 | size: 1 4085 | } 4086 | } 4087 | int_val: -1 4088 | } 4089 | } 4090 | } 4091 | } 4092 | node { 4093 | name: "BoxPredictor_0/ClassPredictor/BiasAdd/Flatten" 4094 | op: "Flatten" 4095 | input: "BoxPredictor_0/ClassPredictor/BiasAdd" 4096 | } 4097 | node { 4098 | name: "BoxPredictor_1/ClassPredictor/BiasAdd/Flatten" 4099 | op: "Flatten" 4100 | input: "BoxPredictor_1/ClassPredictor/BiasAdd" 4101 | } 4102 | node { 4103 | name: "BoxPredictor_2/ClassPredictor/BiasAdd/Flatten" 4104 | op: "Flatten" 4105 | input: "BoxPredictor_2/ClassPredictor/BiasAdd" 4106 | } 4107 | node { 4108 | name: "BoxPredictor_3/ClassPredictor/BiasAdd/Flatten" 4109 | op: "Flatten" 4110 | input: "BoxPredictor_3/ClassPredictor/BiasAdd" 4111 | } 4112 | node { 4113 | name: "BoxPredictor_4/ClassPredictor/BiasAdd/Flatten" 4114 | op: "Flatten" 4115 | input: "BoxPredictor_4/ClassPredictor/BiasAdd" 4116 | } 4117 | node { 4118 | name: "BoxPredictor_5/ClassPredictor/BiasAdd/Flatten" 4119 | op: "Flatten" 4120 | input: "BoxPredictor_5/ClassPredictor/BiasAdd" 4121 | } 4122 | node { 4123 | name: "ClassPredictor/concat" 4124 | op: "ConcatV2" 4125 | input: "BoxPredictor_0/ClassPredictor/BiasAdd/Flatten" 4126 | input: "BoxPredictor_1/ClassPredictor/BiasAdd/Flatten" 4127 | input: "BoxPredictor_2/ClassPredictor/BiasAdd/Flatten" 4128 | input: "BoxPredictor_3/ClassPredictor/BiasAdd/Flatten" 4129 | input: "BoxPredictor_4/ClassPredictor/BiasAdd/Flatten" 4130 | input: "BoxPredictor_5/ClassPredictor/BiasAdd/Flatten" 4131 | input: "concat/axis_flatten" 4132 | } 4133 | node { 4134 | name: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd/Flatten" 4135 | op: "Flatten" 4136 | input: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd" 4137 | } 4138 | node { 4139 | name: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd/Flatten" 4140 | op: "Flatten" 4141 | input: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd" 4142 | } 4143 | node { 4144 | name: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd/Flatten" 4145 | op: "Flatten" 4146 | input: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd" 4147 | } 4148 | node { 4149 | name: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd/Flatten" 4150 | op: "Flatten" 4151 | input: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd" 4152 | } 4153 | node { 4154 | name: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd/Flatten" 4155 | op: "Flatten" 4156 | input: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd" 4157 | } 4158 | node { 4159 | name: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd/Flatten" 4160 | op: "Flatten" 4161 | input: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd" 4162 | } 4163 | node { 4164 | name: "BoxEncodingPredictor/concat" 4165 | op: "ConcatV2" 4166 | input: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd/Flatten" 4167 | input: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd/Flatten" 4168 | input: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd/Flatten" 4169 | input: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd/Flatten" 4170 | input: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd/Flatten" 4171 | input: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd/Flatten" 4172 | input: "concat/axis_flatten" 4173 | } 4174 | node { 4175 | name: "PriorBox_0" 4176 | op: "PriorBox" 4177 | input: "BoxPredictor_0/BoxEncodingPredictor/BiasAdd" 4178 | input: "image_tensor" 4179 | attr { 4180 | key: "clip" 4181 | value { 4182 | b: false 4183 | } 4184 | } 4185 | attr { 4186 | key: "flip" 4187 | value { 4188 | b: false 4189 | } 4190 | } 4191 | attr { 4192 | key: "height" 4193 | value { 4194 | tensor { 4195 | dtype: DT_FLOAT 4196 | tensor_shape { 4197 | dim { 4198 | size: 3 4199 | } 4200 | } 4201 | float_val: 30.0 4202 | float_val: 42.4264068604 4203 | float_val: 84.8528137207 4204 | } 4205 | } 4206 | } 4207 | attr { 4208 | key: "variance" 4209 | value { 4210 | tensor { 4211 | dtype: DT_FLOAT 4212 | tensor_shape { 4213 | dim { 4214 | size: 4 4215 | } 4216 | } 4217 | float_val: 0.10000000149 4218 | float_val: 0.10000000149 4219 | float_val: 0.20000000298 4220 | float_val: 0.20000000298 4221 | } 4222 | } 4223 | } 4224 | attr { 4225 | key: "width" 4226 | value { 4227 | tensor { 4228 | dtype: DT_FLOAT 4229 | tensor_shape { 4230 | dim { 4231 | size: 3 4232 | } 4233 | } 4234 | float_val: 30.0 4235 | float_val: 84.8528137207 4236 | float_val: 42.4264068604 4237 | } 4238 | } 4239 | } 4240 | } 4241 | node { 4242 | name: "PriorBox_1" 4243 | op: "PriorBox" 4244 | input: "BoxPredictor_1/BoxEncodingPredictor/BiasAdd" 4245 | input: "image_tensor" 4246 | attr { 4247 | key: "clip" 4248 | value { 4249 | b: false 4250 | } 4251 | } 4252 | attr { 4253 | key: "flip" 4254 | value { 4255 | b: false 4256 | } 4257 | } 4258 | attr { 4259 | key: "height" 4260 | value { 4261 | tensor { 4262 | dtype: DT_FLOAT 4263 | tensor_shape { 4264 | dim { 4265 | size: 6 4266 | } 4267 | } 4268 | float_val: 105.0 4269 | float_val: 74.2462158203 4270 | float_val: 148.492431641 4271 | float_val: 60.6217765808 4272 | float_val: 181.956329346 4273 | float_val: 125.499000549 4274 | } 4275 | } 4276 | } 4277 | attr { 4278 | key: "variance" 4279 | value { 4280 | tensor { 4281 | dtype: DT_FLOAT 4282 | tensor_shape { 4283 | dim { 4284 | size: 4 4285 | } 4286 | } 4287 | float_val: 0.10000000149 4288 | float_val: 0.10000000149 4289 | float_val: 0.20000000298 4290 | float_val: 0.20000000298 4291 | } 4292 | } 4293 | } 4294 | attr { 4295 | key: "width" 4296 | value { 4297 | tensor { 4298 | dtype: DT_FLOAT 4299 | tensor_shape { 4300 | dim { 4301 | size: 6 4302 | } 4303 | } 4304 | float_val: 105.0 4305 | float_val: 148.492431641 4306 | float_val: 74.2462158203 4307 | float_val: 181.865341187 4308 | float_val: 60.5914611816 4309 | float_val: 125.499000549 4310 | } 4311 | } 4312 | } 4313 | } 4314 | node { 4315 | name: "PriorBox_2" 4316 | op: "PriorBox" 4317 | input: "BoxPredictor_2/BoxEncodingPredictor/BiasAdd" 4318 | input: "image_tensor" 4319 | attr { 4320 | key: "clip" 4321 | value { 4322 | b: false 4323 | } 4324 | } 4325 | attr { 4326 | key: "flip" 4327 | value { 4328 | b: false 4329 | } 4330 | } 4331 | attr { 4332 | key: "height" 4333 | value { 4334 | tensor { 4335 | dtype: DT_FLOAT 4336 | tensor_shape { 4337 | dim { 4338 | size: 6 4339 | } 4340 | } 4341 | float_val: 150.0 4342 | float_val: 106.066017151 4343 | float_val: 212.132034302 4344 | float_val: 86.6025390625 4345 | float_val: 259.93762207 4346 | float_val: 171.026306152 4347 | } 4348 | } 4349 | } 4350 | attr { 4351 | key: "variance" 4352 | value { 4353 | tensor { 4354 | dtype: DT_FLOAT 4355 | tensor_shape { 4356 | dim { 4357 | size: 4 4358 | } 4359 | } 4360 | float_val: 0.10000000149 4361 | float_val: 0.10000000149 4362 | float_val: 0.20000000298 4363 | float_val: 0.20000000298 4364 | } 4365 | } 4366 | } 4367 | attr { 4368 | key: "width" 4369 | value { 4370 | tensor { 4371 | dtype: DT_FLOAT 4372 | tensor_shape { 4373 | dim { 4374 | size: 6 4375 | } 4376 | } 4377 | float_val: 150.0 4378 | float_val: 212.132034302 4379 | float_val: 106.066017151 4380 | float_val: 259.807617188 4381 | float_val: 86.5592269897 4382 | float_val: 171.026306152 4383 | } 4384 | } 4385 | } 4386 | } 4387 | node { 4388 | name: "PriorBox_3" 4389 | op: "PriorBox" 4390 | input: "BoxPredictor_3/BoxEncodingPredictor/BiasAdd" 4391 | input: "image_tensor" 4392 | attr { 4393 | key: "clip" 4394 | value { 4395 | b: false 4396 | } 4397 | } 4398 | attr { 4399 | key: "flip" 4400 | value { 4401 | b: false 4402 | } 4403 | } 4404 | attr { 4405 | key: "height" 4406 | value { 4407 | tensor { 4408 | dtype: DT_FLOAT 4409 | tensor_shape { 4410 | dim { 4411 | size: 6 4412 | } 4413 | } 4414 | float_val: 195.0 4415 | float_val: 137.885818481 4416 | float_val: 275.771636963 4417 | float_val: 112.583305359 4418 | float_val: 337.918914795 4419 | float_val: 216.333084106 4420 | } 4421 | } 4422 | } 4423 | attr { 4424 | key: "variance" 4425 | value { 4426 | tensor { 4427 | dtype: DT_FLOAT 4428 | tensor_shape { 4429 | dim { 4430 | size: 4 4431 | } 4432 | } 4433 | float_val: 0.10000000149 4434 | float_val: 0.10000000149 4435 | float_val: 0.20000000298 4436 | float_val: 0.20000000298 4437 | } 4438 | } 4439 | } 4440 | attr { 4441 | key: "width" 4442 | value { 4443 | tensor { 4444 | dtype: DT_FLOAT 4445 | tensor_shape { 4446 | dim { 4447 | size: 6 4448 | } 4449 | } 4450 | float_val: 195.0 4451 | float_val: 275.771636963 4452 | float_val: 137.885818481 4453 | float_val: 337.749908447 4454 | float_val: 112.527000427 4455 | float_val: 216.333084106 4456 | } 4457 | } 4458 | } 4459 | } 4460 | node { 4461 | name: "PriorBox_4" 4462 | op: "PriorBox" 4463 | input: "BoxPredictor_4/BoxEncodingPredictor/BiasAdd" 4464 | input: "image_tensor" 4465 | attr { 4466 | key: "clip" 4467 | value { 4468 | b: false 4469 | } 4470 | } 4471 | attr { 4472 | key: "flip" 4473 | value { 4474 | b: false 4475 | } 4476 | } 4477 | attr { 4478 | key: "height" 4479 | value { 4480 | tensor { 4481 | dtype: DT_FLOAT 4482 | tensor_shape { 4483 | dim { 4484 | size: 6 4485 | } 4486 | } 4487 | float_val: 240.0 4488 | float_val: 169.705627441 4489 | float_val: 339.411254883 4490 | float_val: 138.564071655 4491 | float_val: 415.90020752 4492 | float_val: 261.533935547 4493 | } 4494 | } 4495 | } 4496 | attr { 4497 | key: "variance" 4498 | value { 4499 | tensor { 4500 | dtype: DT_FLOAT 4501 | tensor_shape { 4502 | dim { 4503 | size: 4 4504 | } 4505 | } 4506 | float_val: 0.10000000149 4507 | float_val: 0.10000000149 4508 | float_val: 0.20000000298 4509 | float_val: 0.20000000298 4510 | } 4511 | } 4512 | } 4513 | attr { 4514 | key: "width" 4515 | value { 4516 | tensor { 4517 | dtype: DT_FLOAT 4518 | tensor_shape { 4519 | dim { 4520 | size: 6 4521 | } 4522 | } 4523 | float_val: 240.0 4524 | float_val: 339.411254883 4525 | float_val: 169.705627441 4526 | float_val: 415.692199707 4527 | float_val: 138.494766235 4528 | float_val: 261.533935547 4529 | } 4530 | } 4531 | } 4532 | } 4533 | node { 4534 | name: "PriorBox_5" 4535 | op: "PriorBox" 4536 | input: "BoxPredictor_5/BoxEncodingPredictor/BiasAdd" 4537 | input: "image_tensor" 4538 | attr { 4539 | key: "clip" 4540 | value { 4541 | b: false 4542 | } 4543 | } 4544 | attr { 4545 | key: "flip" 4546 | value { 4547 | b: false 4548 | } 4549 | } 4550 | attr { 4551 | key: "height" 4552 | value { 4553 | tensor { 4554 | dtype: DT_FLOAT 4555 | tensor_shape { 4556 | dim { 4557 | size: 6 4558 | } 4559 | } 4560 | float_val: 285.0 4561 | float_val: 201.525436401 4562 | float_val: 403.050872803 4563 | float_val: 164.544830322 4564 | float_val: 493.881469727 4565 | float_val: 292.403839111 4566 | } 4567 | } 4568 | } 4569 | attr { 4570 | key: "variance" 4571 | value { 4572 | tensor { 4573 | dtype: DT_FLOAT 4574 | tensor_shape { 4575 | dim { 4576 | size: 4 4577 | } 4578 | } 4579 | float_val: 0.10000000149 4580 | float_val: 0.10000000149 4581 | float_val: 0.20000000298 4582 | float_val: 0.20000000298 4583 | } 4584 | } 4585 | } 4586 | attr { 4587 | key: "width" 4588 | value { 4589 | tensor { 4590 | dtype: DT_FLOAT 4591 | tensor_shape { 4592 | dim { 4593 | size: 6 4594 | } 4595 | } 4596 | float_val: 285.0 4597 | float_val: 403.050872803 4598 | float_val: 201.525436401 4599 | float_val: 493.634490967 4600 | float_val: 164.462539673 4601 | float_val: 292.403839111 4602 | } 4603 | } 4604 | } 4605 | } 4606 | node { 4607 | name: "PriorBox/concat" 4608 | op: "ConcatV2" 4609 | input: "PriorBox_0" 4610 | input: "PriorBox_1" 4611 | input: "PriorBox_2" 4612 | input: "PriorBox_3" 4613 | input: "PriorBox_4" 4614 | input: "PriorBox_5" 4615 | input: "concat/axis_flatten" 4616 | } 4617 | node { 4618 | name: "ClassPredictor/concat/sigmoid" 4619 | op: "Sigmoid" 4620 | input: "ClassPredictor/concat" 4621 | } 4622 | node { 4623 | name: "detection_out" 4624 | op: "DetectionOutput" 4625 | input: "BoxEncodingPredictor/concat" 4626 | input: "ClassPredictor/concat/sigmoid" 4627 | input: "PriorBox/concat" 4628 | attr { 4629 | key: "background_label_id" 4630 | value { 4631 | i: 0 4632 | } 4633 | } 4634 | attr { 4635 | key: "code_type" 4636 | value { 4637 | s: "CENTER_SIZE" 4638 | } 4639 | } 4640 | attr { 4641 | key: "confidence_threshold" 4642 | value { 4643 | f: 0.00999999977648 4644 | } 4645 | } 4646 | attr { 4647 | key: "keep_top_k" 4648 | value { 4649 | i: 100 4650 | } 4651 | } 4652 | attr { 4653 | key: "nms_threshold" 4654 | value { 4655 | f: 0.600000023842 4656 | } 4657 | } 4658 | attr { 4659 | key: "num_classes" 4660 | value { 4661 | i: 91 4662 | } 4663 | } 4664 | attr { 4665 | key: "share_location" 4666 | value { 4667 | b: true 4668 | } 4669 | } 4670 | attr { 4671 | key: "top_k" 4672 | value { 4673 | i: 100 4674 | } 4675 | } 4676 | } 4677 | library { 4678 | } --------------------------------------------------------------------------------