├── README.md ├── __pycache__ ├── assin_priors.cpython-37.pyc ├── basket_utils.cpython-37.pyc ├── dataset.cpython-37.pyc ├── lossfn.cpython-37.pyc ├── model.cpython-37.pyc ├── nms.cpython-37.pyc ├── post_processer.cpython-37.pyc ├── priors.cpython-37.pyc ├── python_nms.cpython-37.pyc └── viz.cpython-37.pyc ├── a.jpg ├── augment ├── __init__.py └── augmentor.py ├── augmentor.py ├── basket_utils.py ├── config.py ├── dataset.py ├── datasets ├── images │ ├── 009814.jpg │ ├── 045346.jpg │ ├── 059925.jpg │ ├── 102954.jpg │ ├── 121349.jpg │ ├── 147664.jpg │ ├── 167843.jpg │ ├── 201586.jpg │ ├── 213702.jpg │ ├── 216026.jpg │ ├── 250583.jpg │ ├── 257559.jpg │ ├── 260655.jpg │ ├── 273406.jpg │ ├── 275153.jpg │ ├── 286036.jpg │ ├── 308634.jpg │ ├── 325869.jpg │ ├── 346044.jpg │ ├── 447131.jpg │ ├── 492718.jpg │ ├── 542925.jpg │ ├── 567307.jpg │ ├── 574125.jpg │ ├── 605272.jpg │ ├── 630439.jpg │ ├── 645593.jpg │ ├── 721074.jpg │ ├── 721205.jpg │ ├── 771259.jpg │ ├── 785985.jpg │ ├── 795439.jpg │ ├── 826272.jpg │ ├── 845712.jpg │ ├── 847258.jpg │ ├── 858721.jpg │ ├── 875555.jpg │ ├── 884419.jpg │ ├── 896922.jpg │ ├── 898712.jpg │ ├── 916700.jpg │ ├── 918904.jpg │ ├── 945782.jpg │ ├── 959677.jpg │ ├── 960859.jpg │ ├── 968216.jpg │ ├── 974557.jpg │ ├── 976494.jpg │ └── 998724.jpg └── labels │ ├── 009814.xml │ ├── 045346.xml │ ├── 059925.xml │ ├── 102954.xml │ ├── 121349.xml │ ├── 147664.xml │ ├── 167843.xml │ ├── 201586.xml │ ├── 213702.xml │ ├── 216026.xml │ ├── 250583.xml │ ├── 257559.xml │ ├── 260655.xml │ ├── 273406.xml │ ├── 275153.xml │ ├── 286036.xml │ ├── 308634.xml │ ├── 325869.xml │ ├── 346044.xml │ ├── 447131.xml │ ├── 492718.xml │ ├── 542925.xml │ ├── 567307.xml │ ├── 574125.xml │ ├── 605272.xml │ ├── 630439.xml │ ├── 645593.xml │ ├── 721074.xml │ ├── 721205.xml │ ├── 771259.xml │ ├── 785985.xml │ ├── 795439.xml │ ├── 826272.xml │ ├── 845712.xml │ ├── 847258.xml │ ├── 858721.xml │ ├── 875555.xml │ ├── 884419.xml │ ├── 896922.xml │ ├── 898712.xml │ ├── 916700.xml │ ├── 918904.xml │ ├── 945782.xml │ ├── 959677.xml │ ├── 960859.xml │ ├── 968216.xml │ ├── 974557.xml │ ├── 976494.xml │ └── 998724.xml ├── demo.png ├── lossfn.py ├── model.py ├── nms.py ├── post_processer.py ├── predict.py ├── priors.py ├── python_nms.py ├── rf_erf_visualize.png ├── split_pic.py ├── train.py ├── visualize_demo.py └── viz.py /README.md: -------------------------------------------------------------------------------- 1 | # BasketNet 2 | 3 | This is a demo of a LFFD model for a Basketball Recognition Competition. 4 | 5 | The demo is based on SSD and LFFD. 6 | ## Install 7 | ```[cmd] 8 | git clone https://github.com/aoru45/LFFD-Pytorch.git 9 | ``` 10 | 11 | ## Usage 12 | Download or make your own dataset and modify the dataset.py file. 13 | ```[cmd] 14 | python train.py 15 | ``` 16 | 17 | ## rf and erf visualize: 18 | ![rf_erf_visualize.png](rf_erf_visualize.png) 19 | 20 | The visulaization code is avaliable here: https://github.com/aoru45/LFFD-Pytorch/blob/master/visualize_demo.py 21 | 22 | ## Network Structure 23 | 24 | ![demo](demo.png) 25 | 26 | 27 | ## demo result 28 | 29 | ![a](a.jpg) 30 | 31 | ## Reference 32 | 33 | SSD:https://arxiv.org/abs/1512.02325 34 | 35 | LFFD:https://arxiv.org/pdf/1904.10633.pdf 36 | 37 | 38 | -------------------------------------------------------------------------------- /__pycache__/assin_priors.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/assin_priors.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/basket_utils.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/basket_utils.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/dataset.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/dataset.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/lossfn.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/lossfn.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/model.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/model.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/nms.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/nms.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/post_processer.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/post_processer.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/priors.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/priors.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/python_nms.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/python_nms.cpython-37.pyc -------------------------------------------------------------------------------- /__pycache__/viz.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/__pycache__/viz.cpython-37.pyc -------------------------------------------------------------------------------- /a.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/a.jpg -------------------------------------------------------------------------------- /augment/__init__.py: -------------------------------------------------------------------------------- 1 | 2 | -------------------------------------------------------------------------------- /augment/augmentor.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | """ 3 | This module provides many types of image augmentation. One can choose appropriate augmentation for 4 | detection, segmentation and classification. 5 | """ 6 | import cv2 7 | import numpy 8 | import random 9 | 10 | 11 | class Augmentor(object): 12 | """ 13 | All augmentation operations are static methods of this class. 14 | """ 15 | 16 | def __init__(self): 17 | pass 18 | 19 | @staticmethod 20 | def histogram_equalisation(image): 21 | """ 22 | do histogram equlisation for grayscale image 23 | :param image: input image with single channel 8bits 24 | :return: processed image 25 | """ 26 | if image.ndim != 2: 27 | print('Input image is not grayscale!') 28 | return None 29 | if image.dtype != numpy.uint8: 30 | print('Input image is not uint8!') 31 | return None 32 | 33 | result = cv2.equalizeHist(image) 34 | return result 35 | 36 | @staticmethod 37 | def grayscale(image): 38 | """ 39 | convert BGR image to grayscale image 40 | :param image: input image with BGR channels 41 | :return: 42 | """ 43 | if image.ndim != 3: 44 | return None 45 | if image.dtype != numpy.uint8: 46 | print('Input image is not uint8!') 47 | return None 48 | 49 | result = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) 50 | return result 51 | 52 | @staticmethod 53 | def inversion(image): 54 | """ 55 | invert the image (255-) 56 | :param image: input image with BGR or grayscale 57 | :return: 58 | """ 59 | if image.dtype != numpy.uint8: 60 | print('Input image is not uint8!') 61 | return None 62 | 63 | result = 255 - image 64 | return result 65 | 66 | @staticmethod 67 | def binarization(image, block_size=5, C=10): 68 | """ 69 | convert input image to binary image 70 | cv2.adaptiveThreshold is used, for detailed information, refer to opencv docs 71 | :param image: 72 | :return: 73 | """ 74 | if image.ndim == 3: 75 | image_grayscale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) 76 | else: 77 | image_grayscale = image 78 | 79 | binary_image = cv2.adaptiveThreshold(image_grayscale, 255, cv2.ADAPTIVE_THRESH_MEAN_C, 80 | cv2.THRESH_BINARY, block_size, C) 81 | return binary_image 82 | 83 | @staticmethod 84 | def brightness(image, min_factor=0.5, max_factor=1.5): 85 | ''' 86 | adjust the image brightness 87 | :param image: 88 | :param min_factor: 89 | :param max_factor: 90 | :return: 91 | ''' 92 | if image.dtype != numpy.uint8: 93 | print('Input image is not uint8!') 94 | return None 95 | 96 | factor = numpy.random.uniform(min_factor, max_factor) 97 | result = image * factor 98 | if factor > 1: 99 | result[result > 255] = 255 100 | result = result.astype(numpy.uint8) 101 | return result 102 | 103 | @staticmethod 104 | def saturation(image, min_factor=0.5, max_factor=1.5): 105 | ''' 106 | adjust the image saturation 107 | :param image: 108 | :param min_factor: 109 | :param max_factor: 110 | :return: 111 | ''' 112 | if image.dtype != numpy.uint8: 113 | print('Input image is not uint8!') 114 | return None 115 | 116 | image_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) 117 | factor = numpy.random.uniform(min_factor, max_factor) 118 | 119 | result = numpy.zeros(image.shape, dtype=numpy.float32) 120 | result[:, :, 0] = image[:, :, 0] * factor + image_gray * (1 - factor) 121 | result[:, :, 1] = image[:, :, 1] * factor + image_gray * (1 - factor) 122 | result[:, :, 2] = image[:, :, 2] * factor + image_gray * (1 - factor) 123 | result[result > 255] = 255 124 | result[result < 0] = 0 125 | result = result.astype(numpy.uint8) 126 | return result 127 | 128 | @staticmethod 129 | def contrast(image, min_factor=0.5, max_factor=1.5): 130 | ''' 131 | adjust the image contrast 132 | :param image: 133 | :param min_factor: 134 | :param max_factor: 135 | :return: 136 | ''' 137 | if image.dtype != numpy.uint8: 138 | print('Input image is not uint8!') 139 | return None 140 | 141 | image_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) 142 | gray_mean = numpy.mean(image_gray) 143 | temp = numpy.ones((image.shape[0], image.shape[1]), dtype=numpy.float32) * gray_mean 144 | factor = numpy.random.uniform(min_factor, max_factor) 145 | 146 | result = numpy.zeros(image.shape, dtype=numpy.float32) 147 | result[:, :, 0] = image[:, :, 0] * factor + temp * (1 - factor) 148 | result[:, :, 1] = image[:, :, 1] * factor + temp * (1 - factor) 149 | result[:, :, 2] = image[:, :, 2] * factor + temp * (1 - factor) 150 | 151 | result[result > 255] = 255 152 | result[result < 0] = 0 153 | result = result.astype(numpy.uint8) 154 | 155 | return result 156 | 157 | @staticmethod 158 | def blur(image, mode='random', kernel_size=3, sigma=1): 159 | """ 160 | 161 | :param image: 162 | :param mode: options 'normalized' 'gaussian' 'median' 163 | :param kernel_size: 164 | :param sigma: used for gaussian blur 165 | :return: 166 | """ 167 | if image.dtype != numpy.uint8: 168 | print('Input image is not uint8!') 169 | return None 170 | 171 | if mode == 'random': 172 | mode = random.choice(['normalized', 'gaussian', 'median']) 173 | 174 | if mode == 'normalized': 175 | result = cv2.blur(image, (kernel_size, kernel_size)) 176 | elif mode == 'gaussian': 177 | result = cv2.GaussianBlur(image, (kernel_size, kernel_size), sigmaX=sigma, sigmaY=sigma) 178 | elif mode == 'median': 179 | result = cv2.medianBlur(image, kernel_size) 180 | else: 181 | print('Blur mode is not supported: %s.' % mode) 182 | result = image 183 | return result 184 | 185 | @staticmethod 186 | def rotation(image, degree=10, mode='crop', scale=1): 187 | """ 188 | 189 | :param image: 190 | :param degree: 191 | :param mode: 'crop'-keep original size, 'fill'-keep full image 192 | :param scale: 193 | :return: 194 | """ 195 | if image.dtype != numpy.uint8: 196 | print('Input image is not uint8!') 197 | return None 198 | 199 | h, w = image.shape[:2] 200 | center_x, center_y = w / 2, h / 2 201 | M = cv2.getRotationMatrix2D((center_x, center_y), degree, scale) 202 | 203 | if mode == 'crop': 204 | new_w, new_h = w, h 205 | else: 206 | cos = numpy.abs(M[0, 0]) 207 | sin = numpy.abs(M[0, 1]) 208 | new_w = int(h * sin + w * cos) 209 | new_h = int(h * cos + w * sin) 210 | M[0, 2] += (new_w / 2) - center_x 211 | M[1, 2] += (new_h / 2) - center_y 212 | 213 | result = cv2.warpAffine(image, M, (new_w, new_h)) 214 | return result 215 | 216 | @staticmethod 217 | def flip(image, orientation='h'): 218 | ''' 219 | 220 | :param image: 221 | :param orientation: 222 | :return: 223 | ''' 224 | if image.dtype != numpy.uint8: 225 | print('Input image is not uint8!') 226 | return None 227 | 228 | if orientation == 'h': 229 | return cv2.flip(image, 1) 230 | elif orientation == 'v': 231 | return cv2.flip(image, 0) 232 | else: 233 | print('Unsupported orientation: %s.' % orientation) 234 | return image 235 | 236 | @staticmethod 237 | def resize(image, size_in_pixel=None, size_in_scale=None): 238 | """ 239 | 240 | :param image: 241 | :param size_in_pixel: tuple (width, height) 242 | :param size_in_scale: tuple (width_scale, height_scale) 243 | :return: 244 | """ 245 | if image.dtype != numpy.uint8: 246 | print('Input image is not uint8!') 247 | return None 248 | 249 | if size_in_pixel is not None: 250 | return cv2.resize(image, size_in_pixel) 251 | elif size_in_scale is not None: 252 | return cv2.resize(image, (0, 0), fx=size_in_scale[0], fy=size_in_scale[1]) 253 | else: 254 | print('size_in_pixel and size_in_scale are both None.') 255 | return image 256 | 257 | @staticmethod 258 | def crop(image, x, y, width, height): 259 | """ 260 | 261 | :param image: 262 | :param x: crop area top-left x coordinate 263 | :param y: crop area top-left y coordinate 264 | :param width: crop area width 265 | :param height: crop area height 266 | :return: 267 | """ 268 | if image.dtype != numpy.uint8: 269 | print('Input image is not uint8!') 270 | return None 271 | 272 | if image.ndim == 3: 273 | return image[y:y + height, x:x + width, :] 274 | else: 275 | return image[y:y + height, x:x + width] 276 | 277 | @staticmethod 278 | def random_crop(image, width, height): 279 | """ 280 | 281 | :param image: 282 | :param width: crop area width 283 | :param height: crop area height 284 | :return: 285 | """ 286 | if image.dtype != numpy.uint8: 287 | print('Input image is not uint8!') 288 | return False, image 289 | 290 | w_interval = image.shape[1] - width 291 | h_interval = image.shape[0] - height 292 | 293 | if image.ndim == 3: 294 | result = numpy.zeros((height, width, 3), dtype=numpy.uint8) 295 | else: 296 | result = numpy.zeros((height, width), dtype=numpy.uint8) 297 | 298 | if w_interval >= 0 and h_interval >= 0: 299 | crop_x, crop_y = random.randint(0, w_interval), random.randint(0, h_interval) 300 | if image.ndim == 3: 301 | result = image[crop_y:crop_y + height, crop_x:crop_x + width, :] 302 | else: 303 | result = image[crop_y:crop_y + height, crop_x:crop_x + width] 304 | elif w_interval < 0 and h_interval >= 0: 305 | put_x = -w_interval / 2 306 | crop_y = random.randint(0, h_interval) 307 | if image.ndim == 3: 308 | result[:, put_x:put_x + image.shape[1], :] = image[crop_y:crop_y + height, :, :] 309 | else: 310 | result[:, put_x:put_x + image.shape[1]] = image[crop_y:crop_y + height, :] 311 | elif w_interval >= 0 and h_interval < 0: 312 | crop_x = random.randint(0, w_interval) 313 | put_y = -h_interval / 2 314 | if image.ndim == 3: 315 | result[put_y:put_y + image.shape[0], :, :] = image[:, crop_x:crop_x + width, :] 316 | else: 317 | result[put_y:put_y + image.shape[0], :] = image[:, crop_x:crop_x + width] 318 | else: 319 | put_x, put_y = -w_interval / 2, -h_interval / 2 320 | if image.ndim == 3: 321 | result[put_y:put_y + image.shape[0], put_x:put_x + image.shape[1], :] = image[:, :, :] 322 | else: 323 | result[put_y:put_y + image.shape[0], put_x:put_x + image.shape[1]] = image[:, :] 324 | 325 | return result 326 | 327 | -------------------------------------------------------------------------------- /augmentor.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-17 00:45:53 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-02 18:30:47 8 | ''' 9 | import cv2 as cv 10 | import numpy as np 11 | from PIL import Image 12 | from albumentations import ( 13 | 14 | HorizontalFlip, 15 | Resize, 16 | RandomSizedBBoxSafeCrop, 17 | Compose, 18 | RandomSunFlare, 19 | RandomShadow, 20 | RandomBrightness, 21 | RandomContrast, 22 | RandomCrop, 23 | GaussianBlur, 24 | CenterCrop, 25 | SmallestMaxSize, 26 | PadIfNeeded, 27 | LongestMaxSize 28 | ) 29 | from torchvision.transforms import ToTensor 30 | class BasketAug(object): 31 | def __init__(self,to_tensor = True): 32 | augs = [ 33 | LongestMaxSize(640), 34 | PadIfNeeded(640,640,cv.BORDER_CONSTANT,value = 0), 35 | #RandomSizedBBoxSafeCrop(height = 300,width = 300), 36 | RandomBrightness(p=0.5), 37 | RandomContrast(p=0.5), 38 | 39 | #RandomSunFlare(p=0.5, flare_roi=(0, 0, 1, 0.5), angle_lower=0.5,src_radius= 150), 40 | RandomShadow(p=0.5, num_shadows_lower=1, num_shadows_upper=1, 41 | shadow_dimension=5, shadow_roi=(0, 0.5, 1, 1)), 42 | HorizontalFlip(p=0.5), 43 | GaussianBlur(p=0.5), 44 | ] 45 | self.transform = Compose(augs, 46 | bbox_params={"format" : "albumentations","min_area": 0,"min_visibility": 0.2,'label_fields': ['category_id']} 47 | ) 48 | 49 | def __call__(self,cv_img, boxes=None, labels=None): 50 | auged = self.transform(image = cv_img,bboxes = boxes, category_id = labels) 51 | return ToTensor()(auged["image"]),auged["bboxes"],auged["category_id"] 52 | if __name__ == "__main__": 53 | basket_aug = BasketAug() 54 | img = cv.imread("./a.jpg") 55 | boxes = [[0.2,0.2,0.5,0.5],[0,0,0.1,0.1]] 56 | labels = [0,1] 57 | img,boxes,labels = basket_aug(img,boxes,labels) 58 | 59 | for box in boxes: 60 | xmin,ymin,xmax,ymax = box 61 | xmin*=640 62 | ymin*=640 63 | xmax*=640 64 | ymax*=640 65 | cv.rectangle(img,(int(xmin),int(ymin)),(int(xmax),int(ymax)),(255,0,0),2) 66 | cv.imshow("img",img) 67 | cv.waitKey(0) 68 | -------------------------------------------------------------------------------- /basket_utils.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-02 01:32:25 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-02 18:37:41 8 | ''' 9 | import torch 10 | import math 11 | from config import * 12 | 13 | def convert_locations_to_boxes(locations, priors, variance = 2): 14 | 15 | # priors can have one dimension less. 16 | if priors.dim() + 1 == locations.dim(): 17 | priors = priors.unsqueeze(0) 18 | return torch.cat([ 19 | priors[..., :2] - locations[..., :2] * priors[..., 2:], 20 | priors[..., :2] - locations[..., 2:] * priors[..., 2:] 21 | ], dim=locations.dim() - 1) 22 | 23 | 24 | def convert_boxes_to_locations(corner_form_boxes, priors,variance = 2): 25 | 26 | if priors.dim() + 1 == corner_form_boxes.dim(): 27 | priors = priors.unsqueeze(0) 28 | return torch.cat([ 29 | (priors[..., :2] - corner_form_boxes[..., :2]) / priors[..., 2:], 30 | (priors[..., :2] - corner_form_boxes[..., 2:]) / priors[..., 2:] 31 | ], dim=corner_form_boxes.dim() - 1) 32 | 33 | def area_of(left_top, right_bottom): 34 | hw = torch.clamp(right_bottom - left_top, min=0.0) 35 | return hw[..., 0] * hw[..., 1] 36 | 37 | 38 | def iou_of(boxes0, boxes1, eps=1e-5): 39 | overlap_left_top = torch.max(boxes0[..., :2], boxes1[..., :2]) # 每个predbox与gt左上角坐标比一下，取最大的，因为是两个坐标之间的比较 40 | overlap_right_bottom = torch.min(boxes0[..., 2:], boxes1[..., 2:]) 41 | 42 | overlap_area = area_of(overlap_left_top, overlap_right_bottom) 43 | area0 = area_of(boxes0[..., :2], boxes0[..., 2:]) 44 | area1 = area_of(boxes1[..., :2], boxes1[..., 2:]) 45 | return overlap_area / (area0 + area1 - overlap_area + eps) 46 | 47 | def in_center(priors,gt): 48 | # 1,n,4 m,1,4 49 | s1 = (priors[...,0] > gt[...,0]) & (priors[...,1] > gt[...,1]) & (priors[...,0] < gt[...,2]) & (priors[...,1] < gt[...,3]) 50 | return s1 51 | ''' 52 | def assign_priors(gt_boxes, gt_labels, priors, 53 | iou_threshold): 54 | 55 | #print(gt_boxes,corner_form_priors) 56 | # size: num_priors x num_targets 57 | # prior 包含 gt 58 | s1 = in_center(priors.unsqueeze(1),gt_boxes.unsqueeze(0)) # 直接匹配在中间的 59 | 60 | #s1[torch.sum(s1,dim = 1) > 1] = False # 同时有多个匹配的 61 | not_ignored = torch.ones(s1.size(0),dtype = torch.uint8) 62 | not_ignored[torch.sum(s1,dim = 1) > 1] = 0 63 | best_target_per_prior, best_target_per_prior_index = s1.max(1) 64 | 65 | 66 | labels = gt_labels[best_target_per_prior_index] # (num_priors,1) 67 | labels[best_target_per_prior == False] = 0 # the backgournd id,没有匹配的给背景 68 | 69 | boxes = gt_boxes[best_target_per_prior_index] #num_priors，４ 70 | t = 0 71 | center_form_gt_boxes = corner_form_to_center_form(boxes) 72 | for f,scale in zip(feature_maps,scales): 73 | d = torch.min(center_form_gt_boxes[t:t+f*f, 2],center_form_gt_boxes[t:t+f*f, 3]) 74 | condition = (d < scale[0]/image_size) | (d > scale[1]/image_size) 75 | labels[t:t+f*f][condition] = 0 76 | 77 | left_gray_scale = [0.9 *scale[0], scale[0]] 78 | right_gray_scale = [scale[1], 1.1 * scale[1]] 79 | 80 | not_ignored[t:t+f*f][(d > left_gray_scale[0]) & (d < left_gray_scale[1])] = 0 81 | not_ignored[t:t+f*f][(d > right_gray_scale[0]) & (d < right_gray_scale[1])] = 0 82 | 83 | t += f*f 84 | return boxes, labels, not_ignored 85 | ''' 86 | def assign_priors(gt_boxes, gt_labels, priors, 87 | iou_threshold): 88 | 89 | #print(gt_boxes,corner_form_priors) 90 | # size: num_priors x num_targets 91 | # prior 包含 gt 92 | s1 = in_center(priors.unsqueeze(1),gt_boxes.unsqueeze(0)) # 直接匹配在中间的 93 | corner_form_priors = center_form_to_corner_form(priors) 94 | ious = iou_of(corner_form_priors.unsqueeze(1),gt_boxes.unsqueeze(0)) 95 | #s1[torch.sum(s1,dim = 1) > 1] = False # 同时有多个匹配的 96 | not_ignored = torch.ones(s1.size(0),dtype = torch.uint8) 97 | not_ignored[torch.sum(s1,dim = 1) > 1] = 0 98 | best_target_per_prior, best_target_per_prior_index = (ious*s1.float()).max(1) 99 | best_prior_per_target, best_prior_per_target_index = ious.max(0) 100 | for target_index, prior_index in enumerate(best_prior_per_target_index): 101 | best_target_per_prior_index[prior_index] = target_index 102 | best_target_per_prior.index_fill_(0, best_prior_per_target_index, 2) 103 | labels = gt_labels[best_target_per_prior_index] # (num_priors,1) 104 | labels[best_target_per_prior < iou_threshold] = 0 # the backgournd id,没有匹配的给背景 105 | 106 | boxes = gt_boxes[best_target_per_prior_index] #num_priors，４ 107 | t = 0 108 | center_form_gt_boxes = corner_form_to_center_form(boxes) 109 | for f,scale in zip(feature_maps,scales): 110 | d = torch.min(center_form_gt_boxes[t:t+f*f, 2],center_form_gt_boxes[t:t+f*f, 3]) 111 | #condition = (d < scale[0]/image_size) | (d > scale[1]/image_size) 112 | #labels[t:t+f*f][condition] = 0 113 | 114 | left_gray_scale = [0.9 *scale[0], scale[0]] 115 | right_gray_scale = [scale[1], 1.1 * scale[1]] 116 | 117 | not_ignored[t:t+f*f][(d > left_gray_scale[0]) & (d < left_gray_scale[1])] = 0 118 | not_ignored[t:t+f*f][(d > right_gray_scale[0]) & (d < right_gray_scale[1])] = 0 119 | 120 | t += f*f 121 | return boxes, labels, not_ignored 122 | 123 | def hard_negative_mining(loss, labels, neg_pos_ratio): 124 | pos_mask = labels > 0 125 | num_pos = pos_mask.long().sum(dim=1, keepdim=True) 126 | num_neg = num_pos * neg_pos_ratio 127 | 128 | loss[pos_mask] = -math.inf 129 | _, indexes = loss.sort(dim=1, descending=True) 130 | _, orders = indexes.sort(dim=1) 131 | neg_mask = orders < num_neg 132 | return pos_mask | neg_mask 133 | 134 | 135 | def center_form_to_corner_form(locations): 136 | return torch.cat([locations[..., :2] - locations[..., 2:] / 2, 137 | locations[..., :2] + locations[..., 2:] / 2], locations.dim() - 1) 138 | 139 | 140 | def corner_form_to_center_form(boxes): 141 | return torch.cat([ 142 | (boxes[..., :2] + boxes[..., 2:]) / 2, 143 | boxes[..., 2:] - boxes[..., :2] 144 | ], boxes.dim() - 1) 145 | -------------------------------------------------------------------------------- /config.py: -------------------------------------------------------------------------------- 1 | num_classes = 3 2 | num_epochs = 500 3 | image_size = 640 4 | strides = [4,4,8,8,16,32,32,32] 5 | feature_maps = [159,159,79,79,39,19,19,19] 6 | scale_factors = [15, 20, 40, 70, 110, 250, 400, 560] 7 | scales = [(10,15),(15,20),(20,40),(40,70),(70,110),(110,250),(250,400),(400,560)] 8 | -------------------------------------------------------------------------------- /dataset.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-10 12:51:20 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-02 18:38:22 8 | ''' 9 | import torch 10 | import torch.nn as nn 11 | from torch.utils.data import Dataset 12 | import glob 13 | from PIL import Image 14 | import xml.etree.ElementTree as ET 15 | import numpy as np 16 | from priors import Priors 17 | from basket_utils import * 18 | np.set_printoptions(threshold=np.inf) 19 | 20 | 21 | class BasketDataset(Dataset): 22 | def __init__(self,img_path,transform = None): 23 | self.img_paths = glob.glob(img_path + "/images/*.jpg") 24 | self.labels = [label.replace(".jpg",".xml").replace("images","labels") for label in self.img_paths] 25 | self.class_names = ("__background__","basketball","volleyball") 26 | prior = Priors() 27 | self.priors = prior() # center form 28 | self.imgW,self.imgH = 640,640 29 | self.transform = transform 30 | def __getitem__(self,idx): 31 | img = Image.open(self.img_paths[idx]).convert("RGB") 32 | label_file = self.labels[idx] 33 | gt_bboxes,gt_classes = self._get_annotation(idx) 34 | 35 | if self.transform: 36 | img,gt_bboxes,gt_classes = self.transform(np.array(img),gt_bboxes,gt_classes) 37 | gt_bboxes = torch.tensor(gt_bboxes) 38 | gt_classes = torch.LongTensor(gt_classes) 39 | 40 | gt_bboxes,gt_classes,ignored = assign_priors(gt_bboxes,gt_classes,self.priors,0.5) 41 | locations = convert_boxes_to_locations(gt_bboxes, self.priors,2) 42 | 43 | return [img,locations,gt_classes,ignored] 44 | def _get_annotation(self,idx): 45 | annotation_file = self.labels[idx] 46 | objects = ET.parse(annotation_file).findall("object") 47 | size = ET.parse(annotation_file).find("size") 48 | boxes = [] 49 | labels = [] 50 | #is_difficult = [] 51 | for obj in objects: 52 | class_name = obj.find('name').text.lower().strip() 53 | bbox = obj.find('bndbox') 54 | x1 = float(bbox.find('xmin').text) - 1 55 | y1 = float(bbox.find('ymin').text) - 1 56 | x2 = float(bbox.find('xmax').text) - 1 57 | y2 = float(bbox.find('ymax').text) - 1 58 | imgW = float(size.find('width').text) 59 | imgH = float(size.find('height').text) 60 | boxes.append([x1/imgW,y1/imgH,x2/imgW,y2/imgH]) 61 | labels.append(self.class_names.index(class_name)) 62 | return boxes,labels 63 | def __len__(self): 64 | return len(self.img_paths) 65 | if __name__ == '__main__': 66 | import random 67 | from augmentor import BasketAug 68 | transform = BasketAug(to_tensor = False) 69 | import cv2 as cv 70 | datset = BasketDataset("../BasketNet_circle/datasets",transform = transform) 71 | img,gt_loc,gt_labels,ignored = datset[random.choice(range(len(datset)))] 72 | #print(ignored.size()) 73 | cv_img = np.array(img) 74 | #h,w,_ = cv_img.shape 75 | cv_img = cv.cvtColor(cv_img,cv.COLOR_RGB2BGR) 76 | priors = datset.priors 77 | idx = (gt_labels > 0) & ignored.bool() 78 | loc = convert_locations_to_boxes(gt_loc,datset.priors,2) 79 | loc = loc[idx] 80 | priors = priors[idx] 81 | label = gt_labels[idx] 82 | print(loc.size()) 83 | for i in range(priors.size(0)): 84 | 85 | x,y,r = priors[i,:] 86 | #print(x,y,r) 87 | x = x.item() * 640 88 | y = y.item() * 640 89 | r = r.item() * 640 90 | print(x,y,r) 91 | cv.circle(cv_img,(int(x),int(y)),int(r),(255,0,0),2) 92 | cv.imshow("cv",cv_img) 93 | cv.waitKey(0) 94 | -------------------------------------------------------------------------------- /datasets/images/009814.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/009814.jpg -------------------------------------------------------------------------------- /datasets/images/045346.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/045346.jpg -------------------------------------------------------------------------------- /datasets/images/059925.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/059925.jpg -------------------------------------------------------------------------------- /datasets/images/102954.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/102954.jpg -------------------------------------------------------------------------------- /datasets/images/121349.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/121349.jpg -------------------------------------------------------------------------------- /datasets/images/147664.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/147664.jpg -------------------------------------------------------------------------------- /datasets/images/167843.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/167843.jpg -------------------------------------------------------------------------------- /datasets/images/201586.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/201586.jpg -------------------------------------------------------------------------------- /datasets/images/213702.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/213702.jpg -------------------------------------------------------------------------------- /datasets/images/216026.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/216026.jpg -------------------------------------------------------------------------------- /datasets/images/250583.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/250583.jpg -------------------------------------------------------------------------------- /datasets/images/257559.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/257559.jpg -------------------------------------------------------------------------------- /datasets/images/260655.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/260655.jpg -------------------------------------------------------------------------------- /datasets/images/273406.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/273406.jpg -------------------------------------------------------------------------------- /datasets/images/275153.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/275153.jpg -------------------------------------------------------------------------------- /datasets/images/286036.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/286036.jpg -------------------------------------------------------------------------------- /datasets/images/308634.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/308634.jpg -------------------------------------------------------------------------------- /datasets/images/325869.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/325869.jpg -------------------------------------------------------------------------------- /datasets/images/346044.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/346044.jpg -------------------------------------------------------------------------------- /datasets/images/447131.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/447131.jpg -------------------------------------------------------------------------------- /datasets/images/492718.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/492718.jpg -------------------------------------------------------------------------------- /datasets/images/542925.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/542925.jpg -------------------------------------------------------------------------------- /datasets/images/567307.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/567307.jpg -------------------------------------------------------------------------------- /datasets/images/574125.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/574125.jpg -------------------------------------------------------------------------------- /datasets/images/605272.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/605272.jpg -------------------------------------------------------------------------------- /datasets/images/630439.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/630439.jpg -------------------------------------------------------------------------------- /datasets/images/645593.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/645593.jpg -------------------------------------------------------------------------------- /datasets/images/721074.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/721074.jpg -------------------------------------------------------------------------------- /datasets/images/721205.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/721205.jpg -------------------------------------------------------------------------------- /datasets/images/771259.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/771259.jpg -------------------------------------------------------------------------------- /datasets/images/785985.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/785985.jpg -------------------------------------------------------------------------------- /datasets/images/795439.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/795439.jpg -------------------------------------------------------------------------------- /datasets/images/826272.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/826272.jpg -------------------------------------------------------------------------------- /datasets/images/845712.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/845712.jpg -------------------------------------------------------------------------------- /datasets/images/847258.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/847258.jpg -------------------------------------------------------------------------------- /datasets/images/858721.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/858721.jpg -------------------------------------------------------------------------------- /datasets/images/875555.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/875555.jpg -------------------------------------------------------------------------------- /datasets/images/884419.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/884419.jpg -------------------------------------------------------------------------------- /datasets/images/896922.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/896922.jpg -------------------------------------------------------------------------------- /datasets/images/898712.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/898712.jpg -------------------------------------------------------------------------------- /datasets/images/916700.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/916700.jpg -------------------------------------------------------------------------------- /datasets/images/918904.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/918904.jpg -------------------------------------------------------------------------------- /datasets/images/945782.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/945782.jpg -------------------------------------------------------------------------------- /datasets/images/959677.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/959677.jpg -------------------------------------------------------------------------------- /datasets/images/960859.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/960859.jpg -------------------------------------------------------------------------------- /datasets/images/968216.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/968216.jpg -------------------------------------------------------------------------------- /datasets/images/974557.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/974557.jpg -------------------------------------------------------------------------------- /datasets/images/976494.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/976494.jpg -------------------------------------------------------------------------------- /datasets/images/998724.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/datasets/images/998724.jpg -------------------------------------------------------------------------------- /datasets/labels/009814.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 009814.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/009814.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 11 21 | 346 22 | 41 23 | 374 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 148 33 | 358 34 | 187 35 | 396 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/045346.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 045346.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/045346.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 102 21 | 440 22 | 160 23 | 500 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 387 33 | 311 34 | 423 35 | 346 36 | 37 | 38 | 39 | volleyball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 273 45 | 343 46 | 308 47 | 381 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/059925.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 059925.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/059925.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 416 21 | 425 22 | 477 23 | 482 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/102954.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 102954.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/102954.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 1 18 | 0 19 | 20 | 98 21 | 462 22 | 163 23 | 512 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 384 33 | 330 34 | 422 35 | 367 36 | 37 | 38 | 39 | volleyball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 271 45 | 362 46 | 307 47 | 403 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/121349.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 121349.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/121349.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 275 21 | 327 22 | 310 23 | 363 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 101 33 | 417 34 | 162 35 | 478 36 | 37 | 38 | 39 | basketball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 389 45 | 293 46 | 426 47 | 332 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/147664.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 147664.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/147664.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 240 21 | 271 22 | 276 23 | 307 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 8 33 | 419 34 | 72 35 | 481 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/167843.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 167843.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/167843.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 210 21 | 417 22 | 270 23 | 477 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 382 33 | 331 34 | 419 35 | 368 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/201586.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 201586.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/201586.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 273 21 | 321 22 | 308 23 | 361 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 104 33 | 409 34 | 162 35 | 474 36 | 37 | 38 | 39 | basketball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 387 45 | 291 46 | 423 47 | 329 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/213702.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 213702.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/213702.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 383 21 | 329 22 | 419 23 | 366 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 1 30 | 0 31 | 32 | 96 33 | 460 34 | 161 35 | 512 36 | 37 | 38 | 39 | volleyball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 269 45 | 364 46 | 305 47 | 402 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/216026.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 216026.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/216026.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 2 21 | 357 22 | 26 23 | 388 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 137 33 | 369 34 | 174 35 | 407 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/250583.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 250583.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/250583.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 6 21 | 343 22 | 45 23 | 382 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 341 33 | 423 34 | 405 35 | 484 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/257559.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 257559.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/257559.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 394 21 | 428 22 | 453 23 | 486 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 62 33 | 320 34 | 100 35 | 359 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/260655.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 260655.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/260655.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 144 21 | 375 22 | 180 23 | 414 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 6 33 | 368 34 | 36 35 | 396 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/273406.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 273406.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/273406.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 360 21 | 441 22 | 420 23 | 502 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/275153.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 275153.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/275153.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 385 21 | 338 22 | 420 23 | 374 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 1 30 | 0 31 | 32 | 98 33 | 468 34 | 166 35 | 512 36 | 37 | 38 | 39 | volleyball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 275 45 | 375 46 | 307 47 | 413 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/286036.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 286036.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/286036.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 131 21 | 365 22 | 169 23 | 400 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/308634.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 308634.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/308634.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 98 21 | 406 22 | 157 23 | 463 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 386 33 | 290 34 | 417 35 | 325 36 | 37 | 38 | 39 | volleyball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 269 45 | 321 46 | 303 47 | 352 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/325869.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 325869.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/325869.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 270 21 | 363 22 | 302 23 | 396 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 384 33 | 328 34 | 418 35 | 365 36 | 37 | 38 | 39 | basketball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 97 45 | 459 46 | 155 47 | 510 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/346044.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 346044.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/346044.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 142 21 | 340 22 | 179 23 | 375 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 7 33 | 322 34 | 34 35 | 351 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/447131.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 447131.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/447131.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 103 21 | 409 22 | 159 23 | 471 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 389 33 | 292 34 | 426 35 | 328 36 | 37 | 38 | 39 | volleyball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 275 45 | 321 46 | 309 47 | 360 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/492718.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 492718.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/492718.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 43 21 | 361 22 | 68 23 | 389 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 180 33 | 377 34 | 212 35 | 411 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/542925.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 542925.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/542925.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 146 21 | 344 22 | 185 23 | 381 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 9 33 | 330 34 | 39 35 | 362 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/567307.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 567307.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/567307.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 12 21 | 336 22 | 50 23 | 377 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 346 33 | 425 34 | 409 35 | 486 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/574125.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 574125.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/574125.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 179 21 | 379 22 | 213 23 | 414 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 41 33 | 357 34 | 73 35 | 389 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/605272.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 605272.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/605272.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 373 21 | 445 22 | 436 23 | 506 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 1 30 | 0 31 | 32 | 1 33 | 355 34 | 33 35 | 392 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/630439.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 630439.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/630439.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 149 21 | 352 22 | 189 23 | 390 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 13 33 | 342 34 | 42 35 | 368 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/645593.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 645593.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/645593.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 40 21 | 353 22 | 69 23 | 387 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 174 33 | 371 34 | 214 35 | 409 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/721074.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 721074.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/721074.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 291 21 | 432 22 | 346 23 | 490 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/721205.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 721205.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/721205.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 141 21 | 339 22 | 176 23 | 374 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 1 30 | 0 31 | 32 | 1 33 | 320 34 | 33 35 | 350 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/771259.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 771259.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/771259.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 151 21 | 346 22 | 190 23 | 389 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 12 33 | 326 34 | 48 35 | 359 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/785985.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 785985.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/785985.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 128 21 | 382 22 | 170 23 | 420 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/795439.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 795439.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/795439.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 102 21 | 419 22 | 166 23 | 478 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 388 33 | 298 34 | 428 35 | 332 36 | 37 | 38 | 39 | volleyball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 277 45 | 331 46 | 312 47 | 365 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/826272.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 826272.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/826272.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 277 21 | 330 22 | 315 23 | 368 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 390 33 | 296 34 | 430 35 | 332 36 | 37 | 38 | 39 | basketball 40 | Unspecified 41 | 0 42 | 0 43 | 44 | 105 45 | 425 46 | 165 47 | 483 48 | 49 | 50 | 51 | -------------------------------------------------------------------------------- /datasets/labels/845712.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 845712.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/845712.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 147 21 | 308 22 | 180 23 | 341 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/847258.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 847258.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/847258.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 7 21 | 346 22 | 48 23 | 386 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 345 33 | 425 34 | 408 35 | 485 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/858721.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 858721.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/858721.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 86 21 | 332 22 | 122 23 | 372 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 1 30 | 0 31 | 32 | 463 33 | 445 34 | 512 35 | 503 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/875555.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 875555.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/875555.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 143 21 | 344 22 | 184 23 | 380 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 7 33 | 326 34 | 39 35 | 356 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/884419.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 884419.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/884419.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 24 21 | 344 22 | 64 23 | 380 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 407 33 | 446 34 | 468 35 | 507 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/896922.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 896922.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/896922.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 390 21 | 128 22 | 432 23 | 180 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/898712.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 898712.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/898712.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 140 21 | 374 22 | 180 23 | 416 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 3 33 | 361 34 | 29 35 | 393 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/916700.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 916700.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/916700.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 56 21 | 428 22 | 124 23 | 490 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/918904.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 918904.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/918904.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 139 21 | 342 22 | 180 23 | 377 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 3 33 | 327 34 | 33 35 | 360 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/945782.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 945782.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/945782.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 13 21 | 332 22 | 41 23 | 363 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 148 33 | 350 34 | 187 35 | 391 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/959677.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 959677.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/959677.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 195 21 | 322 22 | 247 23 | 359 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 83 33 | 345 34 | 121 35 | 379 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/960859.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 960859.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/960859.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 348 21 | 425 22 | 405 23 | 485 24 | 25 | 26 | 27 | basketball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 8 33 | 341 34 | 45 35 | 378 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/968216.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 968216.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/968216.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 44 21 | 333 22 | 83 23 | 368 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/974557.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 974557.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/974557.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 169 21 | 368 22 | 212 23 | 409 24 | 25 | 26 | 27 | volleyball 28 | Unspecified 29 | 0 30 | 0 31 | 32 | 35 33 | 355 34 | 68 35 | 384 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /datasets/labels/976494.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 976494.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/976494.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | volleyball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 137 21 | 280 22 | 173 23 | 314 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /datasets/labels/998724.xml: -------------------------------------------------------------------------------- 1 | 2 | images 3 | 998724.jpg 4 | /media/xueaoru/其他/ML/my-SSD/dataset/images/998724.jpg 5 | 6 | Unknown 7 | 8 | 9 | 512 10 | 512 11 | 3 12 | 13 | 0 14 | 15 | basketball 16 | Unspecified 17 | 0 18 | 0 19 | 20 | 126 21 | 363 22 | 164 23 | 400 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /demo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/demo.png -------------------------------------------------------------------------------- /lossfn.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-10 00:03:42 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-01 23:59:09 8 | ''' 9 | import torch 10 | import torch.nn as nn 11 | import torch.nn.functional as F 12 | import math 13 | import basket_utils 14 | 15 | class BasketLoss(nn.Module): 16 | def __init__(self): 17 | super(BasketLoss, self).__init__() 18 | self.neg_pos_ratio = 10 19 | def forward(self,scores, predicted_locations, labels, gt_locations,not_ignored): 20 | num_classes = scores.size(2) 21 | with torch.no_grad(): 22 | loss = -F.log_softmax(scores, dim=2)[:, :, 0] 23 | mask = basket_utils.hard_negative_mining(loss, labels, self.neg_pos_ratio) & not_ignored 24 | 25 | confidence = scores[mask, :] 26 | classification_loss = F.cross_entropy(confidence.view(-1, num_classes), labels[mask], reduction='sum') 27 | 28 | pos_mask = labels > 0 29 | predicted_locations = predicted_locations[pos_mask, :].view(-1, 4) 30 | gt_locations = gt_locations[pos_mask, :].view(-1, 4) 31 | mse_loss = F.mse_loss(predicted_locations, gt_locations, reduction='sum') 32 | #smooth_l1_loss = F.smooth_l1_loss(predicted_locations, gt_locations, reduction='sum') 33 | num_pos = gt_locations.size(0) 34 | return mse_loss / num_pos, classification_loss / num_pos 35 | -------------------------------------------------------------------------------- /model.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-09 23:13:46 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-02 18:39:48 8 | ''' 9 | 10 | 11 | import torch 12 | import torch.nn as nn 13 | import torch.nn.functional as F 14 | from priors import Priors 15 | from basket_utils import * 16 | class SeparableConv2d(nn.Module): 17 | def __init__(self,in_channels,out_channels,kernel_size=3,stride=1,padding=1): 18 | super(SeparableConv2d, self).__init__() 19 | self.conv = nn.Sequential( 20 | nn.Conv2d(in_channels, in_channels, kernel_size = kernel_size, stride=stride, groups=in_channels,padding=padding), 21 | nn.Conv2d(in_channels, out_channels, kernel_size = 1, stride=(1, 1)) 22 | ) 23 | def forward(self,x): 24 | return self.conv(x) 25 | class ResBlock(nn.Module): 26 | def __init__(self,channels): 27 | super(ResBlock, self).__init__() 28 | self.conv2dRelu = nn.Sequential( 29 | SeparableConv2d(channels,channels,kernel_size=3,stride=1,padding=1), 30 | nn.ReLU6(channels), 31 | SeparableConv2d(channels,channels,kernel_size=3,stride=1,padding=1), 32 | nn.ReLU6(channels) 33 | ) 34 | self.relu = nn.ReLU6(channels) 35 | def forward(self,x): 36 | return self.relu(x + self.conv2dRelu(x)) 37 | class BasketLossBranch(nn.Module): 38 | def __init__(self,in_channels,out_channels=64,num_classes=3): 39 | super(BasketLossBranch, self).__init__() 40 | self.conv1x1relu = nn.Sequential( 41 | nn.Conv2d(in_channels,out_channels,kernel_size=1,stride=1,padding=0), 42 | nn.ReLU6(out_channels) 43 | ) 44 | self.score =nn.Sequential( 45 | nn.Conv2d(out_channels,out_channels,kernel_size=1,stride=1,padding=0), 46 | nn.ReLU6(out_channels), 47 | nn.Conv2d(out_channels,num_classes,kernel_size=1,stride=1) 48 | ) 49 | self.locations =nn.Sequential( 50 | nn.Conv2d(out_channels,out_channels,kernel_size=1,stride=1,padding=0), 51 | nn.ReLU6(out_channels), 52 | nn.Conv2d(out_channels,4,kernel_size=1,stride=1) 53 | ) 54 | def forward(self,x): 55 | score = self.score(self.conv1x1relu(x)) 56 | locations = self.locations(self.conv1x1relu(x)) 57 | return score,locations 58 | class BasketNet(nn.Module): 59 | def __init__(self,num_classes = 3): 60 | super(BasketNet, self).__init__() 61 | self.num_classes = num_classes 62 | self.priors = None 63 | self.c1 = nn.Sequential( 64 | SeparableConv2d(3,64,kernel_size=3,stride=2,padding=0), 65 | nn.ReLU6(64) 66 | ) 67 | self.c2 = nn.Sequential( 68 | SeparableConv2d(64,64,kernel_size=3,stride=2,padding=0), 69 | nn.ReLU6(64) 70 | ) 71 | self.tinypart1 = nn.Sequential( 72 | ResBlock(64), 73 | ResBlock(64), 74 | ResBlock(64) 75 | ) 76 | self.tinypart2 = ResBlock(64) 77 | self.c11 = nn.Sequential( 78 | nn.Conv2d(64,64,kernel_size=3,stride=2,padding=0), 79 | nn.ReLU6(64) 80 | ) 81 | self.smallpart1 = ResBlock(64) 82 | self.smallpart2 = ResBlock(64) 83 | self.c16 = nn.Sequential( 84 | nn.Conv2d(64,128,kernel_size=3,stride=2,padding=0), 85 | nn.ReLU6(128) 86 | ) 87 | self.mediumpart = ResBlock(128) 88 | self.c19 = nn.Sequential( 89 | nn.Conv2d(128,128,kernel_size=3,stride=2,padding=0), 90 | nn.ReLU6(128) 91 | ) 92 | self.largepart1 = ResBlock(128) 93 | self.largepart2 = ResBlock(128) 94 | self.largepart3 = ResBlock(128) 95 | 96 | 97 | self.lossbranch1 = BasketLossBranch(64,num_classes = self.num_classes) 98 | self.lossbranch2 = BasketLossBranch(64,num_classes = self.num_classes) 99 | self.lossbranch3 = BasketLossBranch(64,num_classes = self.num_classes) 100 | self.lossbranch4 = BasketLossBranch(64,num_classes = self.num_classes) 101 | self.lossbranch5 = BasketLossBranch(128,num_classes = self.num_classes) 102 | self.lossbranch6 = BasketLossBranch(128,num_classes = self.num_classes) 103 | self.lossbranch7 = BasketLossBranch(128,num_classes = self.num_classes) 104 | self.lossbranch8 = BasketLossBranch(128,num_classes = self.num_classes) 105 | def forward(self, x): 106 | c1 = self.c1(x) 107 | c2 = self.c2(c1) 108 | 109 | c8 = self.tinypart1(c2) 110 | c10 = self.tinypart2(c8) 111 | 112 | c11 = self.c11(c10) 113 | c13 = self.smallpart1(c11) 114 | c15 = self.smallpart2(c13) 115 | 116 | c16 = self.c16(c15) 117 | c18 = self.mediumpart(c16) 118 | 119 | c19 = self.c19(c18) 120 | c21 = self.largepart1(c19) 121 | c23 = self.largepart2(c21) 122 | c25 = self.largepart3(c23) 123 | 124 | score1,loc1 = self.lossbranch1(c8) 125 | score2,loc2 = self.lossbranch2(c10) 126 | score3,loc3 = self.lossbranch3(c13) 127 | score4,loc4 = self.lossbranch4(c15) 128 | score5,loc5 = self.lossbranch5(c18) 129 | #print(loc1.size(),loc2.size(),loc3.size(),loc4.size(),loc5.size()) 130 | score6,loc6 = self.lossbranch6(c21) 131 | score7,loc7 = self.lossbranch7(c23) 132 | score8,loc8 = self.lossbranch8(c25) 133 | 134 | cls = torch.cat([score1.permute(0, 2, 3, 1).contiguous().view(score1.size(0),-1, self.num_classes), 135 | score2.permute(0, 2, 3, 1).contiguous().view(score2.size(0),-1, self.num_classes), 136 | score3.permute(0, 2, 3, 1).contiguous().view(score3.size(0), -1, self.num_classes), 137 | score4.permute(0, 2, 3, 1).contiguous().view(score4.size(0), -1, self.num_classes), 138 | score5.permute(0, 2, 3, 1).contiguous().view(score5.size(0), -1, self.num_classes), 139 | score6.permute(0, 2, 3, 1).contiguous().view(score6.size(0), -1, self.num_classes), 140 | score7.permute(0, 2, 3, 1).contiguous().view(score7.size(0), -1, self.num_classes), 141 | score8.permute(0, 2, 3, 1).contiguous().view(score8.size(0), -1, self.num_classes)], dim=1) 142 | loc = torch.cat([loc1.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4), 143 | loc2.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4), 144 | loc3.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4), 145 | loc4.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4), 146 | loc5.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4), 147 | loc6.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4), 148 | loc7.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4), 149 | loc8.permute(0, 2, 3, 1).contiguous().view(loc1.size(0), -1,4)], dim=1) 150 | 151 | if not self.training: 152 | if self.priors is None: 153 | self.priors = Priors()() # center form 154 | #self.priors = self.priors.cuda() 155 | boxes = convert_locations_to_boxes( 156 | loc, self.priors, 2 157 | )# corner_form 158 | cls = F.softmax(cls, dim=2) 159 | return cls, boxes 160 | else: 161 | #print(confidences.size(),locations.size()) 162 | return (cls,loc) # (2,1111,3) (2,1111,4) 163 | 164 | 165 | from torchsummary import summary 166 | if __name__ == '__main__': 167 | model = BasketNet() 168 | summary(model,(3,640,640),device = "cpu") 169 | -------------------------------------------------------------------------------- /nms.py: -------------------------------------------------------------------------------- 1 | try: 2 | import torch_extension 3 | 4 | _nms = torch_extension.nms 5 | except ImportError: 6 | from python_nms import python_nms 7 | 8 | _nms = python_nms 9 | 10 | 11 | def boxes_nms(boxes, scores, nms_thresh, max_count=-1): 12 | """ Performs non-maximum suppression, run on GPU or CPU according to 13 | boxes's device. 14 | Args: 15 | boxes(Tensor): `xyxy` mode boxes, use absolute coordinates(not support relative coordinates), 16 | shape is (n, 4) 17 | scores(Tensor): scores, shape is (n, ) 18 | nms_thresh(float): thresh 19 | max_count (int): if > 0, then only the top max_proposals are kept after non-maximum suppression 20 | Returns: 21 | indices kept. 22 | """ 23 | keep = _nms(boxes, scores, nms_thresh) 24 | if max_count > 0: 25 | keep = keep[:max_count] 26 | return keep 27 | -------------------------------------------------------------------------------- /post_processer.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-13 21:06:46 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-03 00:23:31 8 | ''' 9 | import torch 10 | 11 | from nms import boxes_nms 12 | 13 | 14 | class PostProcessor: 15 | def __init__(self, 16 | iou_threshold = 0.10, 17 | score_threshold = 0.8, 18 | image_size = 640, 19 | max_per_class=200, 20 | max_per_image=-1): 21 | self.confidence_threshold = score_threshold 22 | self.iou_threshold = iou_threshold 23 | self.width = image_size 24 | self.height = image_size 25 | self.max_per_class = max_per_class 26 | self.max_per_image = max_per_image 27 | 28 | def __call__(self, confidences, locations, width=None, height=None, batch_ids=None): 29 | if width is None: 30 | width = self.width 31 | if height is None: 32 | height = self.height 33 | 34 | batch_size = confidences.size(0) 35 | if batch_ids is None: 36 | batch_ids = torch.arange(batch_size, device=confidences.device) 37 | else: 38 | batch_ids = torch.tensor(batch_ids, device=confidences.device) 39 | 40 | locations = locations[batch_ids] 41 | confidences = confidences[batch_ids] 42 | 43 | results = [] 44 | for decoded_boxes, scores in zip(locations, confidences): 45 | # per batch 46 | filtered_boxes = [] 47 | filtered_labels = [] 48 | filtered_probs = [] 49 | for class_index in range(1, scores.size(1)): 50 | probs = scores[:, class_index] 51 | mask = probs > self.confidence_threshold 52 | probs = probs[mask] 53 | if probs.size(0) == 0: 54 | continue 55 | boxes = decoded_boxes[mask, :] # x1,y1,x2,y2 56 | #ratio = (width + height) / 2. 57 | boxes[:, 0] *= width 58 | boxes[:, 2] *= width 59 | boxes[:, 1] *= height 60 | boxes[:, 3] *= height 61 | 62 | keep = boxes_nms(boxes, probs, self.iou_threshold, self.max_per_class) 63 | 64 | boxes = boxes[keep, :] 65 | labels = torch.tensor([class_index] * keep.size(0)) 66 | probs = probs[keep] 67 | 68 | filtered_boxes.append(boxes) 69 | filtered_labels.append(labels) 70 | filtered_probs.append(probs) 71 | 72 | # no object detected 73 | if len(filtered_boxes) == 0: 74 | filtered_boxes = torch.empty(0, 4) 75 | filtered_labels = torch.empty(0) 76 | filtered_probs = torch.empty(0) 77 | else: # cat all result 78 | filtered_boxes = torch.cat(filtered_boxes, 0) 79 | filtered_labels = torch.cat(filtered_labels, 0) 80 | filtered_probs = torch.cat(filtered_probs, 0) 81 | if 0 < self.max_per_image < filtered_probs.size(0): 82 | keep = torch.argsort(filtered_probs, dim=0, descending=True)[:self.max_per_image] 83 | filtered_boxes = filtered_boxes[keep, :] 84 | filtered_labels = filtered_labels[keep] 85 | filtered_probs = filtered_probs[keep] 86 | results.append((filtered_boxes, filtered_labels, filtered_probs)) 87 | return results 88 | -------------------------------------------------------------------------------- /predict.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-02 21:08:56 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-03 00:22:04 8 | ''' 9 | import torch 10 | import torchvision 11 | from model import BasketNet 12 | from torchvision import transforms 13 | #from transforms import * 14 | from PIL import Image 15 | import numpy as np 16 | from viz import draw_bounding_boxes,draw_circles 17 | from post_processer import PostProcessor 18 | from priors import Priors 19 | 20 | prior = Priors() 21 | center_form_priors = prior() 22 | 23 | post_process = PostProcessor() 24 | color_map= { 25 | 1: (0,0,255), 26 | 2: (0,255,255) 27 | } 28 | transform = transforms.Compose([ 29 | transforms.Resize((640,640)), 30 | transforms.ToTensor(), 31 | #transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) 32 | ] 33 | ) 34 | 35 | 36 | def pic_test(): 37 | img = Image.open("./test2.jpg").convert('RGB') 38 | image = np.array(img,dtype = np.float32) 39 | height, width, _ = image.shape 40 | img = transform(img) 41 | img = img.unsqueeze(0) 42 | #img = img.cuda() 43 | net = BasketNet() 44 | 45 | net.load_state_dict(torch.load("./ckpt/1748.pth",map_location="cpu")) 46 | #net.cuda() 47 | print("network load...") 48 | net.eval() 49 | with torch.no_grad(): 50 | pred_confidence,boxes = net(img) 51 | 52 | 53 | output = post_process(pred_confidence,boxes, width=width, height=height)[0] 54 | #print(output) 55 | boxes, labels, scores = [o.to("cpu").numpy() for o in output] 56 | 57 | 58 | drawn_image = draw_bounding_boxes(image, boxes, labels, scores, ("__background__","basketball","volleyball")).astype(np.uint8) 59 | 60 | Image.fromarray(drawn_image).save("./a.jpg") 61 | 62 | def cap_test(): 63 | import cv2 as cv 64 | cap = cv.VideoCapture("../test.mp4") 65 | net = BasketNet() 66 | 67 | net.load_state_dict(torch.load("./ckpt/1748.pth",map_location="cpu")) 68 | #net.cuda() 69 | net.eval() 70 | while True: 71 | ret,frame = cap.read() 72 | if not ret: 73 | break 74 | 75 | height,width,_ = frame.shape 76 | center_width = width//2 77 | frame = frame[:,center_width-height//2:center_width + height//2] 78 | height,width,_ = frame.shape 79 | cv_img = cv.cvtColor(frame,cv.COLOR_BGR2RGB) 80 | img = Image.fromarray(cv_img) 81 | img = transform(img) 82 | img = img.unsqueeze(0) 83 | #img = img.cuda() 84 | 85 | with torch.no_grad(): 86 | pred_confidence,boxes = net(img) 87 | 88 | output = post_process(pred_confidence,boxes, width=width, height=height)[0] 89 | boxes, labels, scores = [o.to("cpu").numpy() for o in output] 90 | drawn_image = draw_bounding_boxes(frame, boxes, labels, scores, ("__background__","basketball","volleyball")).astype(np.uint8) 91 | 92 | cv.imshow("img",drawn_image) 93 | key = cv.waitKey(1) 94 | if key == ord("q"): 95 | break 96 | cv.destroyAllWindows() 97 | cap.release() 98 | if __name__ == "__main__": 99 | pic_test() 100 | -------------------------------------------------------------------------------- /priors.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-10 01:02:36 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-02 18:36:33 8 | ''' 9 | from itertools import product 10 | 11 | import torch 12 | from math import sqrt 13 | from config import * 14 | 15 | class Priors: 16 | def __init__(self,clip = True): 17 | self.image_size = image_size 18 | self.strides = strides 19 | self.feature_maps = feature_maps 20 | self.clip = clip 21 | def __call__(self): 22 | priors = [] 23 | for k, f in enumerate(self.feature_maps): 24 | # 513/4 = 128.25 25 | # 126/128.25 = 0.98245 26 | # 126.5/128.25 = 0.98635 27 | scale = self.image_size / self.strides[k] 28 | for i, j in product(range(f), repeat=2): 29 | 30 | cx = (j + 0.5) / scale 31 | cy = (i + 0.5) / scale 32 | scale_factor = scale_factors[k] * 0.5 / self.image_size 33 | #r = self.sizes[k] 34 | #r = r/self.image_size 35 | #h = w = self.sizes[k] / self.image_size 36 | priors.append([cx, cy, scale_factor]) 37 | 38 | priors = torch.tensor(priors) #(num_priors,4) 39 | if self.clip: 40 | priors.clamp_(max=1, min=0) 41 | return priors 42 | def get_priors(): 43 | return Priors() 44 | if __name__ == "__main__": 45 | priors = Priors() 46 | print(priors().size()) 47 | -------------------------------------------------------------------------------- /python_nms.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import numpy as np 3 | 4 | 5 | def python_nms(boxes, scores, nms_thresh, max_count=-1): 6 | """ Performs non-maximum suppression using numpy 7 | Args: 8 | boxes(Tensor): `xyxy` mode boxes, use absolute coordinates(not support relative coordinates), 9 | shape is (n, 4) 10 | scores(Tensor): scores, shape is (n, ) 11 | nms_thresh(float): thresh 12 | max_count (int): if > 0, then only the top max_proposals are kept after non-maximum suppression 13 | Returns: 14 | indices kept. 15 | """ 16 | if boxes.numel() == 0: 17 | return torch.empty((0,), dtype=torch.long) 18 | # Use numpy to run nms. Running nms in PyTorch code on CPU is really slow. 19 | origin_device = boxes.device 20 | cpu_device = torch.device('cpu') 21 | boxes = boxes.to(cpu_device).numpy() 22 | scores = scores.to(cpu_device).numpy() 23 | 24 | x1 = boxes[:, 0] 25 | y1 = boxes[:, 1] 26 | x2 = boxes[:, 2] 27 | y2 = boxes[:, 3] 28 | areas = (x2 - x1 + 1) * (y2 - y1 + 1) 29 | order = np.argsort(scores)[::-1] 30 | num_detections = boxes.shape[0] 31 | suppressed = np.zeros((num_detections,), dtype=np.bool) 32 | for _i in range(num_detections): 33 | i = order[_i] 34 | if suppressed[i]: 35 | continue 36 | ix1 = x1[i] 37 | iy1 = y1[i] 38 | ix2 = x2[i] 39 | iy2 = y2[i] 40 | iarea = areas[i] 41 | 42 | for _j in range(_i + 1, num_detections): 43 | j = order[_j] 44 | if suppressed[j]: 45 | continue 46 | 47 | xx1 = max(ix1, x1[j]) 48 | yy1 = max(iy1, y1[j]) 49 | xx2 = min(ix2, x2[j]) 50 | yy2 = min(iy2, y2[j]) 51 | w = max(0, xx2 - xx1 + 1) 52 | h = max(0, yy2 - yy1 + 1) 53 | 54 | inter = w * h 55 | ovr = inter / (iarea + areas[j] - inter) 56 | if ovr >= nms_thresh: 57 | suppressed[j] = True 58 | keep = np.nonzero(suppressed == 0)[0] 59 | if max_count > 0: 60 | keep = keep[:max_count] 61 | keep = torch.from_numpy(keep).to(origin_device) 62 | return keep 63 | -------------------------------------------------------------------------------- /rf_erf_visualize.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aoru45/LFFD-Pytorch/b76596b7af32e2fddad78275ccd90c09ac8e20bb/rf_erf_visualize.png -------------------------------------------------------------------------------- /split_pic.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-10 18:22:23 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-09-10 18:29:11 8 | ''' 9 | import cv2 as cv 10 | import random 11 | 12 | if __name__ == "__main__": 13 | video_path = "/home/xueaoru/视频/VID_20180815_142126.mp4" 14 | cap = cv.VideoCapture(video_path) 15 | idx = 0 16 | while True: 17 | ret,frame = cap.read() 18 | if not ret: 19 | break 20 | idx +=1 21 | if idx % 30 ==0: 22 | cv.imwrite("./images/{:0>6d}.jpg".format(random.randint(0,999999)),frame) 23 | cv.destroyAllWindows() 24 | 25 | -------------------------------------------------------------------------------- /train.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-10 13:56:50 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-02 20:14:08 8 | ''' 9 | import torch 10 | from torch.utils.data import DataLoader 11 | import torch.nn as nn 12 | import torch.optim as optim 13 | from model import BasketNet 14 | from torchvision import transforms 15 | from dataset import BasketDataset 16 | from lossfn import BasketLoss 17 | from tqdm import tqdm 18 | from augmentor import BasketAug 19 | from config import * 20 | 21 | if __name__ == '__main__': 22 | transform = BasketAug() 23 | #transform = transforms.Compose([ 24 | # transforms.Resize(512), 25 | # transforms.ToTensor(), 26 | # transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) 27 | #]) 28 | dataset = BasketDataset("../BasketNet_circle/datasets",transform = transform) 29 | dataloader = DataLoader(dataset,batch_size = 16,shuffle = True,num_workers = 6,pin_memory=True) 30 | 31 | net = BasketNet(num_classes = num_classes) 32 | net.cuda() 33 | #net.load_state_dict(torch.load("./ckpt/3090.pth")) 34 | optimizer = optim.Adam(net.parameters(),lr = 1e-3) 35 | #optimizer = optim.SGD(net.parameters(), lr=1e-1,momentum=0.9,weight_decay=0.) 36 | #scheduler = optim.lr_scheduler.StepLR(optimizer,100,0.1) 37 | loss_fn = BasketLoss() 38 | num_batches = len(dataloader) 39 | min_loss = float("inf") 40 | for epoch in range(num_epochs): 41 | epoch_loss_cls = 0. 42 | epoch_loss_reg = 0. 43 | 44 | for img,gt_pos,gt_labels,not_ignored in tqdm(dataloader): 45 | img = img.cuda() 46 | gt_pos =gt_pos.cuda() 47 | not_ignored = not_ignored.cuda() 48 | gt_labels =gt_labels.cuda() 49 | cls,loc = net(img) 50 | reg_loss,cls_loss = loss_fn(cls,loc,gt_labels,gt_pos,not_ignored) 51 | epoch_loss_cls += cls_loss.item() 52 | epoch_loss_reg += reg_loss.item() 53 | loss = (reg_loss + cls_loss) 54 | optimizer.zero_grad() 55 | loss.backward() 56 | optimizer.step() 57 | #scheduler.step() 58 | print("cls_loss:{}---reg_loss:{}".format(epoch_loss_cls/num_batches,epoch_loss_reg/num_batches)) 59 | if (epoch_loss_cls/num_batches + epoch_loss_reg/num_batches) < min_loss: 60 | min_loss = epoch_loss_cls/num_batches + epoch_loss_reg/num_batches 61 | torch.save(net.state_dict(),"./ckpt/{}.pth".format(int(epoch_loss_cls/num_batches * 1000 + epoch_loss_reg/num_batches * 1000))) 62 | -------------------------------------------------------------------------------- /visualize_demo.py: -------------------------------------------------------------------------------- 1 | import cv2 as cv 2 | import torch 3 | import torch.nn as nn 4 | import torchvision 5 | from torchvision.models import ResNet 6 | import numpy as np 7 | class ResBlock(nn.Module): 8 | def __init__(self,channels): 9 | super(ResBlock, self).__init__() 10 | self.conv2dRelu = nn.Sequential( 11 | nn.Conv2d(channels,channels,kernel_size=3,stride=1,padding=1), 12 | nn.ReLU(channels), 13 | nn.Conv2d(channels,channels,kernel_size=3,stride=1,padding=1), 14 | nn.ReLU(channels) 15 | ) 16 | self.relu = nn.ReLU(channels) 17 | def forward(self,x): 18 | return self.relu(x + self.conv2dRelu(x)) 19 | class TinyNet(nn.Module): 20 | def __init__(self): 21 | super(TinyNet,self).__init__() 22 | self.c1 = nn.Sequential( 23 | nn.Conv2d(3,64,kernel_size=3,stride=2,padding=0), 24 | nn.ReLU(64) 25 | ) 26 | self.c2 = nn.Sequential( 27 | nn.Conv2d(64,64,kernel_size=3,stride=2,padding=0), 28 | nn.ReLU(64) 29 | ) 30 | self.tinypart1 = nn.Sequential( 31 | ResBlock(64), 32 | ResBlock(64), 33 | ResBlock(64) 34 | ) 35 | def forward(self,x): 36 | c1 = self.c1(x) 37 | c2 = self.c2(c1) 38 | c8 = self.tinypart1(c2) 39 | return c8 40 | if __name__ == "__main__": 41 | model = TinyNet() 42 | for module in model.modules(): 43 | try: 44 | nn.init.constant_(module.weight, 0.05) 45 | nn.init.zeros_(module.bias) 46 | nn.init.zeros_(module.running_mean) 47 | nn.init.ones_(module.running_var) 48 | except Exception as e: 49 | pass 50 | if type(module) is nn.BatchNorm2d: 51 | module.eval() 52 | x = torch.ones(1,3,640,640,requires_grad= True) 53 | pred = model(x) 54 | grad = torch.zeros_like(pred, requires_grad= True) 55 | grad[0, 0, 64, 64] = 1 56 | pred.backward(gradient = grad) 57 | grad_input = x.grad[0,0,...].data.numpy() 58 | grad_input = grad_input / np.max(grad_input) 59 | # 有效感受野 0.75 - 0.85 60 | #grad_input = np.where(grad_input>0.85,1,0) 61 | grad_input = np.where(grad_input>0.75,1,0) 62 | # 注释掉即为感受野 63 | grad_input = (grad_input * 255).astype(np.uint8) 64 | kernel = np.ones((5, 5), np.uint8) 65 | grad_input = cv.dilate(grad_input, kernel=kernel) 66 | 67 | contours, _ = cv.findContours(grad_input, mode=cv.RETR_EXTERNAL, method=cv.CHAIN_APPROX_SIMPLE) 68 | rect = cv.boundingRect(contours[0]) 69 | print(rect[-2:]) 70 | cv.imshow( "a",grad_input) 71 | cv.waitKey(0) -------------------------------------------------------------------------------- /viz.py: -------------------------------------------------------------------------------- 1 | ''' 2 | @Descripttion: This is Aoru Xue's demo,which is only for reference 3 | @version: 4 | @Author: Aoru Xue 5 | @Date: 2019-09-13 21:06:56 6 | @LastEditors: Aoru Xue 7 | @LastEditTime: 2019-10-02 17:00:40 8 | ''' 9 | import numpy as np 10 | from six.moves import range 11 | import PIL.Image as Image 12 | import PIL.ImageDraw as ImageDraw 13 | import PIL.ImageFont as ImageFont 14 | 15 | STANDARD_COLORS = [ 16 | 'AliceBlue', 'Chartreuse', 'Aqua', 'Aquamarine', 'Azure', 'Beige', 'Bisque', 17 | 'BlanchedAlmond', 'BlueViolet', 'BurlyWood', 'CadetBlue', 'AntiqueWhite', 18 | 'Chocolate', 'Coral', 'CornflowerBlue', 'Cornsilk', 'Crimson', 'Cyan', 19 | 'DarkCyan', 'DarkGoldenRod', 'DarkGrey', 'DarkKhaki', 'DarkOrange', 20 | 'DarkOrchid', 'DarkSalmon', 'DarkSeaGreen', 'DarkTurquoise', 'DarkViolet', 21 | 'DeepPink', 'DeepSkyBlue', 'DodgerBlue', 'FireBrick', 'FloralWhite', 22 | 'ForestGreen', 'Fuchsia', 'Gainsboro', 'GhostWhite', 'Gold', 'GoldenRod', 23 | 'Salmon', 'Tan', 'HoneyDew', 'HotPink', 'IndianRed', 'Ivory', 'Khaki', 24 | 'Lavender', 'LavenderBlush', 'LawnGreen', 'LemonChiffon', 'LightBlue', 25 | 'LightCoral', 'LightCyan', 'LightGoldenRodYellow', 'LightGray', 'LightGrey', 26 | 'LightGreen', 'LightPink', 'LightSalmon', 'LightSeaGreen', 'LightSkyBlue', 27 | 'LightSlateGray', 'LightSlateGrey', 'LightSteelBlue', 'LightYellow', 'Lime', 28 | 'LimeGreen', 'Linen', 'Magenta', 'MediumAquaMarine', 'MediumOrchid', 29 | 'MediumPurple', 'MediumSeaGreen', 'MediumSlateBlue', 'MediumSpringGreen', 30 | 'MediumTurquoise', 'MediumVioletRed', 'MintCream', 'MistyRose', 'Moccasin', 31 | 'NavajoWhite', 'OldLace', 'Olive', 'OliveDrab', 'Orange', 'OrangeRed', 32 | 'Orchid', 'PaleGoldenRod', 'PaleGreen', 'PaleTurquoise', 'PaleVioletRed', 33 | 'PapayaWhip', 'PeachPuff', 'Peru', 'Pink', 'Plum', 'PowderBlue', 'Purple', 34 | 'Red', 'RosyBrown', 'RoyalBlue', 'SaddleBrown', 'Green', 'SandyBrown', 35 | 'SeaGreen', 'SeaShell', 'Sienna', 'Silver', 'SkyBlue', 'SlateBlue', 36 | 'SlateGray', 'SlateGrey', 'Snow', 'SpringGreen', 'SteelBlue', 'GreenYellow', 37 | 'Teal', 'Thistle', 'Tomato', 'Turquoise', 'Violet', 'Wheat', 'White', 38 | 'WhiteSmoke', 'Yellow', 'YellowGreen' 39 | ] 40 | 41 | NUM_COLORS = len(STANDARD_COLORS) 42 | 43 | try: 44 | FONT = ImageFont.truetype('arial.ttf', 24) 45 | except IOError: 46 | FONT = ImageFont.load_default() 47 | 48 | 49 | def _draw_single_box(image, xmin, ymin, xmax, ymax, color='black', display_str=None, font=None, thickness=2): 50 | draw = ImageDraw.Draw(image) 51 | left, right, top, bottom = xmin, xmax, ymin, ymax 52 | draw.line([(left, top), (left, bottom), (right, bottom), (right, top), (left, top)], width=thickness, fill=color) 53 | if display_str is not None: 54 | text_bottom = bottom 55 | # Reverse list and print from bottom to top. 56 | text_width, text_height = font.getsize(display_str) 57 | margin = np.ceil(0.05 * text_height) 58 | draw.rectangle( 59 | [(left, text_bottom - text_height - 2 * margin), (left + text_width, text_bottom)], fill=color) 60 | draw.text((left + margin, text_bottom - text_height - margin), 61 | display_str, 62 | fill='black', 63 | font=font) 64 | 65 | return image 66 | 67 | def _draw_single_circle(image, xmin, ymin, xmax, ymax, color='black', display_str=None, font=None, thickness=2): 68 | draw = ImageDraw.Draw(image) 69 | left, right, top, bottom = xmin, xmax, ymin, ymax 70 | draw.ellipse([(left, top), (right, bottom)], width=thickness) 71 | if display_str is not None: 72 | text_bottom = bottom 73 | text_width, text_height = font.getsize(display_str) 74 | margin = np.ceil(0.05 * text_height) 75 | draw.text((left + margin, text_bottom - text_height - margin), 76 | display_str, 77 | fill='black', 78 | font=font) 79 | # text_bottom = bottom 80 | # Reverse list and print from bottom to top. 81 | # text_width, text_height = font.getsize(display_str) 82 | # margin = np.ceil(0.05 * text_height) 83 | #draw.rectangle( 84 | # [(left, text_bottom - text_height - 2 * margin), (left + text_width, text_bottom)], fill=color) 85 | #draw.text((left + margin, text_bottom - text_height - margin), 86 | # display_str, 87 | # fill='black', 88 | # font=font) 89 | 90 | return image 91 | def draw_circles(image, circles, labels=None, probs=None, class_name_map=None): 92 | num_circles = circles.shape[0] 93 | gt_boxes_new = circles.copy() 94 | draw_image = Image.fromarray(np.uint8(image)) 95 | for i in range(num_circles): 96 | display_str = None 97 | this_class = 0 98 | if labels is not None: 99 | this_class = labels[i] 100 | class_name = class_name_map[this_class] if class_name_map is not None else str(this_class) 101 | class_name = class_name.decode('utf-8') if isinstance(class_name, bytes) else class_name 102 | if probs is not None: 103 | prob = probs[i] 104 | display_str = '{}:{:.2f}'.format(class_name, prob) 105 | else: 106 | display_str = class_name 107 | draw_image = _draw_single_circle(image=draw_image, 108 | xmin=gt_boxes_new[i, 0] - gt_boxes_new[i, 2], 109 | ymin=gt_boxes_new[i, 1] - gt_boxes_new[i, 2], 110 | xmax=gt_boxes_new[i, 0] + gt_boxes_new[i, 2], 111 | ymax=gt_boxes_new[i, 1] + gt_boxes_new[i, 2], 112 | color=STANDARD_COLORS[this_class % NUM_COLORS], 113 | display_str=display_str, 114 | font=FONT) 115 | 116 | image = np.array(draw_image, dtype=np.float32) 117 | return image 118 | 119 | 120 | def draw_bounding_boxes(image, boxes, labels=None, probs=None, class_name_map=None): 121 | """Draw bboxes(labels, probs) on image 122 | Args: 123 | image: numpy array image, shape should be (height, width, channel) 124 | boxes: bboxes, shape should be (N, 4), and each row is (xmin, ymin, xmax, ymax) 125 | labels: labels, shape: (N, ) 126 | probs: label scores, shape: (N, ), can be False/True or None 127 | class_name_map: list or dict, map class id to class name for visualization. 128 | can be False/True or None 129 | Returns: 130 | An image with information drawn on it. 131 | """ 132 | num_boxes = boxes.shape[0] 133 | gt_boxes_new = boxes.copy() 134 | draw_image = Image.fromarray(np.uint8(image)) 135 | for i in range(num_boxes): 136 | display_str = None 137 | this_class = 0 138 | if labels is not None: 139 | this_class = labels[i] 140 | class_name = class_name_map[this_class] if class_name_map is not None else str(this_class) 141 | class_name = class_name.decode('utf-8') if isinstance(class_name, bytes) else class_name 142 | if probs is not None: 143 | prob = probs[i] 144 | display_str = '{}:{:.2f}'.format(class_name, prob) 145 | else: 146 | display_str = class_name 147 | draw_image = _draw_single_box(image=draw_image, 148 | xmin=gt_boxes_new[i, 0], 149 | ymin=gt_boxes_new[i, 1], 150 | xmax=gt_boxes_new[i, 2], 151 | ymax=gt_boxes_new[i, 3], 152 | color=STANDARD_COLORS[this_class % NUM_COLORS], 153 | display_str=display_str, 154 | font=FONT) 155 | 156 | image = np.array(draw_image, dtype=np.float32) 157 | return image 158 | --------------------------------------------------------------------------------