def main(): parser = argparse.ArgumentParser() parser.add_argument('--weights', default='yolo_v2.ckpt', type=str) # darknet-19.ckpt parser.add_argument('--weight_dir', default='output', type=str) parser.add_argument('--data_dir', default='data', type=str) parser.add_argument('--gpu', default='', type=str) # which gpu to be selected args = parser.parse_args() os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu # configure gpu weights_file = os.path.join(args.data_dir, args.weight_dir, args.weights) yolo = yolo_v2(False) # 'False' mean 'test' # yolo = Darknet19(False) detector = Detector(yolo, weights_file) #detect the video #cap = cv2.VideoCapture('asd.mp4') #cap = cv2.VideoCapture(0) #detector.video_detect(cap) #detect the image imagename = './test/01.jpg' detector.image_detect(imagename)
def main(): start0 = time.clock() parser = argparse.ArgumentParser() parser.add_argument('--weights', default='yolo_v2.ckpt-4000', type=str) # darknet-19.ckpt parser.add_argument('--weight_dir', default='output', type=str) parser.add_argument('--data_dir', default='data', type=str) parser.add_argument('--gpu', default='0', type=str) # which gpu to be selected args = parser.parse_args() os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu # configure gpu weights_file = os.path.join(args.data_dir, args.weight_dir, args.weights) # yolo_obj = yolo_v2(False) # 'False' mean 'test' yolo_obj = yolo_v2(True) # 单张图片预测,设置True效果好 detector = Detector(yolo_obj, weights_file) elapsed = (time.clock() - start0) print("Time used:", elapsed) #detect the video #cap = cv2.VideoCapture('asd.mp4') #cap = cv2.VideoCapture(0) #detector.video_detect(cap) #detect the image start1 = time.clock() file_path = os.getcwd() + r'/data/data_set/defect_data/train_image' for i in os.listdir(file_path): imagename = os.path.join(file_path, i) detector.image_detect(imagename) elapsed1 = (time.clock() - start1) print("Time used:", elapsed1)
def main(): parser = argparse.ArgumentParser() parser.add_argument( '--weights', default='D:\\reference\\5-dataset\\yolo2_coco_weights\\yolo2_coco.ckpt', type=str) # darknet-19.ckpt parser.add_argument('--gpu', default='0', type=str) # which gpu to be selected args = parser.parse_args() if args.gpu is not None: cfg.GPU = args.gpu if args.weights is not None: cfg.WEIGHTS_FILE = args.weights os.environ['CUDA_VISIBLE_DEVICES'] = cfg.GPU yolo = yolo_v2() # yolo = Darknet19() pre_data = Pascal_voc() train = Train(yolo, pre_data) print('start training ...') train.train() print('successful training.')
def main(): parser = argparse.ArgumentParser() parser.add_argument('--weights', default='yolo_v2.ckpt-4000', type=str) # darknet-19.ckpt # parser.add_argument('--weights', default = '', type = str) # darknet-19.ckpt #voc的预训练权重 # parser.add_argument('--weights', default = 'yolo_weights.ckpt', type = str) parser.add_argument('--gpu', default="0", type=str) # which gpu to be selected args = parser.parse_args() if args.gpu is not None: cfg.GPU = args.gpu if args.weights is not None: cfg.WEIGHTS_FILE = args.weights os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" # os.environ['CUDA_VISIBLE_DEVICES'] = '1,0' os.environ['CUDA_VISIBLE_DEVICES'] = cfg.GPU yolo_obj = yolo_v2() pre_data = Data_preprocess() train = Train(yolo_obj, pre_data) print('start training ...') train.train() print('successful training.')
def main(): parser = argparse.ArgumentParser() # 创建一个解析器对象,并告诉它将会有些什么参数 parser.add_argument('--weights', default = 'yolo_v2.ckpt', type = str) # darknet-19.ckpt parser.add_argument('--gpu', default = '', type = str) # 可以使用的GPU args = parser.parse_args() if args.gpu is not None: cfg.GPU = args.gpu if args.weights is not None: cfg.WEIGHTS_FILE = args.weights os.environ['CUDA_VISIBLE_DEVICES'] = cfg.GPU yolo = yolo_v2() # yolo = Darknet19() pre_data = Pascal_voc() #数据集类 train = Train(yolo, pre_data) print('start training ...') train.train() print('successful training.')
def main(): parser = argparse.ArgumentParser() parser.add_argument('-w', '--weights', default=None, type=str) # darknet-19.ckpt parser.add_argument('--weight_dir', default='output', type=str) parser.add_argument('--data_dir', default='data', type=str) parser.add_argument('-o', '--optimizer', default=1, type=int) parser.add_argument('-v', '--var_set', default='all', type=str, choices=['all', 'back']) parser.add_argument('-g', '--gpu', default='', type=str) # which gpu to be selected args = parser.parse_args() random.seed(cfg.RANDOM_SEED) np.random.seed(cfg.RANDOM_SEED) tf.set_random_seed(cfg.RANDOM_SEED) if args.gpu is not None: cfg.GPU = args.gpu if args.weights is not None: cfg.WEIGHTS_FILE = args.weights else: w_dir = (os.path.join(cfg.DATA_DIR, args.data_dir)) latest = tf.train.latest_checkpoint(w_dir) if latest is not None and len(latest) > 0: cfg.WEIGHTS_FILE = latest print("** resore weights file:", cfg.WEIGHTS_FILE) os.environ['CUDA_VISIBLE_DEVICES'] = cfg.GPU yolo = yolo_v2() # yolo = Darknet19() pre_data2012 = Pascal_voc_VOC2012() pre_data2007 = Pascal_voc_VOC2007() pre_data = pre_data2012.takeIn(pre_data2007) train = Train(yolo, pre_data, optimizer_no=args.optimizer, var_set=args.var_set) print('** start training ...') train.train() print('** successful training.')
def main(): parser = argparse.ArgumentParser() parser.add_argument('--weights', default='yolo_v2.ckpt', type=str) parser.add_argument('--gpu', default='', type=str) args = parser.parse_args() if args.gpu is not None: cfg.GPU = args.gpu if args.weights is not None: cfg.WEIGHTS_FILE = args.weights os.environ['CUDA_VISIBLE_DEVICES'] = cfg.GPU yolo = yolo_v2() pre_data = Pascal_voc() train = Train(yolo, pre_data) print('start training ...') train.train() print('successful training.')
def main(): parser = argparse.ArgumentParser() #创建一个解析器 parser.add_argument('--weights', default = 'yolo_v2.ckpt', type = str) # darknet-19.ckpt parser.add_argument('--weight_dir', default = 'output', type = str) parser.add_argument('--data_dir', default = 'data', type = str) parser.add_argument('--gpu', default = '', type = str) # which gpu to be selected args = parser.parse_args()# 解析输入的命令行,参数默认是从sys.argv[1:]中获取,parse_args()返回一个命名空间 # 包含传递给命令行的参数,该对象将参数保存为属性 os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu # 可见GPU weights_file = os.path.join(args.data_dir, args.weight_dir, args.weights) yolo = yolo_v2(False) # 'False' mean 'test' # yolo = Darknet19(False) detector = Detector(yolo, weights_file) #detect the video #cap = cv2.VideoCapture('asd.mp4') #cap = cv2.VideoCapture(0) #detector.video_detect(cap) #detect the image imagename = './test/02.jpg' detector.image_detect(imagename)
def main(): parser = argparse.ArgumentParser() parser.add_argument('--weights', default = 'yolo_v2.ckpt-30000', type = str) # darknet-19.ckpt parser.add_argument('--gpu', default = '', type = str) # which gpu to be selected args = parser.parse_args() if args.gpu is not None: cfg.GPU = args.gpu if args.weights is not None: cfg.WEIGHTS_FILE = args.weights os.environ['CUDA_VISIBLE_DEVICES'] = '2'#cfg.GPU yolo = yolo_v2() # yolo = Darknet19() # pre_data = Pascal_voc() AIZOO = AIZOO_dataset() train = Train(yolo, AIZOO) print('start training ...') train.train() print('successful training.')
def main(): parser = argparse.ArgumentParser() parser.add_argument('--weights', default='yolo_v2.ckpt', type=str) # darknet-19.ckpt parser.add_argument('--weight_dir', default='output', type=str) parser.add_argument('--data_dir', default='data', type=str) parser.add_argument('--gpu', default='', type=str) # which gpu to be selected args = parser.parse_args() os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu # configure gpu weights_file = os.path.join(args.data_dir, args.weight_dir, args.weights) yolo = yolo_v2(False) # 'False' mean 'test' # yolo = Darknet19(False) detector = Detector(yolo, weights_file) #detect the image image_files_path = './linemod/cfg/test_shuf_labels.txt' #imagename = './test/02.jpg' #detector.image_detect(imagename) detector.test(image_files_path)
def main(): parser = argparse.ArgumentParser() parser.add_argument('-i', '--images', nargs='+', type=str, required=True) parser.add_argument('--weights', default=None, type=str) # darknet-19.ckpt parser.add_argument('--weight_dir', default='output', type=str) parser.add_argument('--data_dir', default='data', type=str) parser.add_argument('--gpu', default='', type=str) # which gpu to be selected args = parser.parse_args() os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu # configure gpu weights_dir = os.path.join(args.data_dir, args.weight_dir) if args.weights is not None: cfg.WEIGHTS_FILE = args.weights else: latest = tf.train.latest_checkpoint(weights_dir) if latest is not None and len(latest) > 0: cfg.WEIGHTS_FILE = os.path.basename(latest) weights_file = os.path.join(weights_dir, cfg.WEIGHTS_FILE) print("using weigts file:", weights_file) yolo = yolo_v2(False) # 'False' mean 'test' # yolo = Darknet19(False) detector = Detector(yolo, weights_file) #detect the video #cap = cv2.VideoCapture('asd.mp4') #cap = cv2.VideoCapture(0) #detector.video_detect(cap) #detect the image for imagename in args.images: key = detector.image_detect(imagename) print('hitted key=', key) if key == 27 or key == 1048603: break # means ESC via SSH
break cap.release() cv2.destroyAllWindows() def main(): parser = argparse.ArgumentParser() parser.add_argument('--weights', default = 'yolo_v2.ckpt', type = str) # darknet-19.ckpt parser.add_argument('--weight_dir', default = 'output', type = str) parser.add_argument('--data_dir', default = 'data', type = str) parser.add_argument('--gpu', default = '', type = str) # which gpu to be selected args = parser.parse_args() os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu # configure gpu weights_file = os.path.join(args.data_dir, args.weight_dir, args.weights) yolo = yolo_v2(False) # 'False' mean 'test' # yolo = Darknet19(False) detector = Detector(yolo, weights_file) #detect the video #cap = cv2.VideoCapture('asd.mp4') #cap = cv2.VideoCapture(0) #detector.video_detect(cap) #detect the image imagename = './test/01.jpg' detector.image_detect(imagename) if __name__ == '__main__': main()