def read_data(self, list_of_videos): for vid in list_of_videos: file_ptr = open(vid, 'r') content = file_ptr.read().split('\n')[:-1] obs_perc = [.1, .2, .3, .5] for i in range(len(obs_perc)): observed_content = content[:int(obs_perc[i]*len(content))] input_vid = encode_content(observed_content, self.nRows, self.nCols, self.actions_dict) input_vid = np.reshape(input_vid, [self.nRows, self.nCols, 1]) target_content = content[int(obs_perc[i]*len(content)):int((0.5+obs_perc[i])*len(content))] target = encode_content(target_content, self.nRows, self.nCols, self.actions_dict) target = np.reshape(target, [self.nRows, self.nCols, 1]) example = [input_vid, target] self.list_of_examples.append(example) random.shuffle(self.list_of_examples) return
observed_content=[] vid_len = 0 if args.input_type == "gt": file_ptr = open(vid, 'r') content = file_ptr.read().split('\n')[:-1] vid_len = len(content) for obs_p in obs_percentages: if args.input_type == "decoded": file_ptr = open(args.decoded_path+"/obs"+str(obs_p)+"/"+f_name+'.txt', 'r') observed_content = file_ptr.read().split('\n')[:-1] vid_len = int(len(observed_content)/obs_p) elif args.input_type == "gt": observed_content = content[:int(obs_p*vid_len)] input_x = encode_content(observed_content, args.nRows, nClasses, actions_dict) input_x = [np.reshape(input_x, [args.nRows, nClasses, 1])] with tf.Session() as sess: label_seq, length_seq = model.predict(sess, model_restore_path, input_x, args.sigma, actions_dict) recognition = [] for i in range(len(label_seq)): recognition = np.concatenate((recognition, [label_seq[i]]*int(0.5*vid_len*length_seq[i]/args.nRows))) recognition = np.concatenate((observed_content,recognition)) diff = int((0.5+obs_p)*vid_len)-len(recognition) for i in range(diff): recognition = np.concatenate((recognition, [label_seq[-1]])) #write results to file for pred_p in pred_percentages: path=args.results_save_path+"/obs"+str(obs_p)+"-pred"+str(pred_p)