def main(): log.basicConfig(format="[ %(levelname)s ] %(message)s", level=log.INFO, stream=sys.stdout) args = build_argparser().parse_args() cfg = load_module(args.config) exec_net, plugin, input_blob, out_blob, shape = load_ir_model( args.model, args.device, args.plugin_dir, args.cpu_extension) n_batch, channels, height, width = shape image = cv2.imread(args.input_image, cv2.IMREAD_GRAYSCALE) img_to_display = image.copy() cv2.imshow('sample image', image) print("press key to run inference engine") cv2.waitKey(0) # waits until a key is pressed in_frame = cv2.resize(image, (width, height)) print("in_frame shape", in_frame.shape) #in_frame = in_frame.transpose((2, 0, 1)) # Change data layout from HWC to CHW in_frame = in_frame.reshape((n_batch, channels, height, width)) #start = time.time() #for i in range(10000): result = exec_net.infer(inputs={input_blob: in_frame}) #end = time.time() #print("elapsed time is ms: ", (end - start)/10) gazeX_out = result[out_blob][0] gazeY_out = result[out_blob][1] print("gazeX, gazeY out: ") print(gazeX_out, gazeY_out) del exec_net del plugin
def main(): log.basicConfig(format="[ %(levelname)s ] %(message)s", level=log.INFO, stream=sys.stdout) args = build_argparser().parse_args() cfg = load_module(args.config) exec_net, plugin, input_blob, out_blob, shape = load_ir_model( args.model, args.device, args.plugin_dir, args.cpu_extension) n_batch, channels, height, width = shape image = cv2.imread(args.input_image) img_to_display = image.copy() in_frame = cv2.resize(image, (width, height)) in_frame = in_frame.transpose( (2, 0, 1)) # Change data layout from HWC to CHW in_frame = in_frame.reshape((n_batch, channels, height, width)) result = exec_net.infer(inputs={input_blob: in_frame}) lp_code = result[out_blob][0] lp_number = decode_ie_output(lp_code, cfg.r_vocab) print('Output: {}'.format(lp_number)) img_to_display = display_license_plate(lp_number, img_to_display) cv2.imshow('License Plate', img_to_display) cv2.waitKey(0) del exec_net del plugin
def main(): args = build_argparser().parse_args() graph = load_graph(args.model) config = load_module(args.config) image = cv2.imread(args.input_image) img = cv2.resize(image, (94, 24)) img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) img = np.float32(img) img = np.multiply(img, 1.0 / 255.0) input = graph.get_tensor_by_name("import/input:0") output = graph.get_tensor_by_name("import/d_predictions:0") with tf.Session(graph=graph) as sess: results = sess.run(output, feed_dict={input: [img]}) print(results) decoded_lp = decode_beams(results, config.r_vocab)[0] print(decoded_lp) img_to_display = display_license_plate(decoded_lp, image) cv2.imshow('License Plate', img_to_display) cv2.waitKey(0)
def main(_): args = parse_args() config = load_module(args.path_to_config) checkpoint = args.checkpoint if args.checkpoint else tf.train.latest_checkpoint(config.model_dir) print(checkpoint) if not checkpoint or not os.path.isfile(checkpoint+'.index'): raise FileNotFoundError(str(checkpoint)) step = checkpoint.split('.')[-2].split('-')[-1] output_dir = args.output_dir if args.output_dir else os.path.join(config.model_dir, 'export_{}'.format(step)) # Freezing graph frozen_dir = os.path.join(output_dir, 'frozen_graph') frozen_graph = freezing_graph(config, checkpoint, frozen_dir) # Export to IR export_dir = os.path.join(output_dir, 'IR', args.data_type) mo_params = { 'framework': 'tf', 'model_name': 'lpr', 'input': 'input', 'output': 'd_predictions', 'reverse_input_channels': True, 'scale': 255, 'input_shape': [1] + list(config.input_shape), 'data_type': args.data_type, } execute_mo(mo_params, frozen_graph, export_dir)
def main(_): args = parse_args() config = load_module(args.path_to_config) checkpoint = args.checkpoint if args.checkpoint else tf.train.latest_checkpoint( config.MODEL_DIR) print(checkpoint) if not checkpoint or not os.path.isfile(checkpoint + '.index'): raise FileNotFoundError(str(checkpoint)) step = checkpoint.split('-')[-1] output_dir = args.output_dir if args.output_dir else os.path.join( config.MODEL_DIR, 'export_{}'.format(step)) # Freezing graph frozen_dir = os.path.join(output_dir, 'frozen_graph') frozen_graph, ssd_config_path, train_param, ssd_config = freezing_graph( config, checkpoint, frozen_dir) # Export to IR export_dir = os.path.join(output_dir, 'IR', args.data_type) input_shape = [1] + list(config.input_shape) # Add batch size 1 in shape scale = 1. / train_param.scale mean_value = [train_param.mean_value for _ in range(3)] mo_params = { 'model_name': args.model_name, 'output': ','.join(ssd_config['cut_points']), 'input_shape': input_shape, 'scale': scale, 'mean_value': mean_value, 'tensorflow_use_custom_operations_config': ssd_config_path, 'data_type': args.data_type, } execute_mo(mo_params, frozen_graph, export_dir)
def main(_): args = parse_args() cfg = load_module(args.path_to_config) infer(cfg, args.input_type, args.input, args.conf_threshold, args.dump_to_json, args.show, args.dump_output_video, args.path_to_output_video)
def main(_): args = parse_args() config = load_module(args.path_to_config) checkpoint = args.checkpoint if args.checkpoint else tf.train.latest_checkpoint( config.model_dir) print("PATH OF CHECKPOINT: {}:".format(checkpoint)) if not checkpoint or not os.path.isfile(checkpoint + '.index'): raise FileNotFoundError(str(checkpoint)) step = checkpoint.split('.')[-2].split('-')[-1] output_dir = args.output_dir if args.output_dir else os.path.join( config.model_dir, 'export_{}'.format(step)) # Freezing graph frozen_dir = os.path.join(output_dir, 'frozen_graph') frozen_graph = freezing_graph(config, checkpoint, frozen_dir)
def main(_): args = parse_args() cfg = load_module(args.path_to_config) session_config = create_session(cfg, 'eval') run_config = tf.estimator.RunConfig(session_config=session_config) va_estimator = tf.estimator.Estimator( model_fn=resnet_v1_10_1, params=cfg.resnet_params, model_dir=cfg.model_dir, config=run_config) eval_data = InputEvalData(batch_size=cfg.eval.batch_size, input_shape=cfg.input_shape, json_path=cfg.eval.annotation_path) eval_loop(va_estimator, eval_data, cfg)
def main(_): args = parse_args() cfg = load_module(args.path_to_config) train(cfg, args.init_checkpoint)
def main(_): args = parse_args() cfg = load_module(args.path_to_config) eval_loop(cfg)
def main(_): args = parse_args() cfg = load_module(args.path_to_config) validate(cfg)
def main(_): args = parse_args() cfg = load_module(args.path_to_config) infer(cfg)