actionformer-master.zip

{'dataset': {'crop_ratio': [0.9, 1.0], 'default_fps': None, 'downsample_rate': 1, 'feat_folder': './data/74/res2net50', 'feat_stride': 1, 'file_ext': '.npy', 'file_prefix': None, 'force_upsampling': False, 'input_dim': 2048, 'json_file': './data/74/74_anno_score_4myres2net.json', 'max_seq_len': 4608, 'num_classes': 32, 'num_frames': 1, 'trunc_thresh': 0.5}, 'dataset_name': 'medicine', 'devices': ['cuda:0'], 'init_rand_seed': 1234567891, 'loader': {'batch_size': 1, 'num_workers': 1}, 'model': {'backbone_arch': (2, 2, 5), 'backbone_type': 'convTransformer', 'embd_dim': 512, 'embd_kernel_size': 3, 'embd_with_ln': True, 'fpn_dim': 512, 'fpn_start_level': 0, 'fpn_type': 'identity', 'fpn_with_ln': True, 'head_dim': 512, 'head_kernel_size': 3, 'head_num_layers': 3, 'head_with_ln': True, 'input_dim': 2048, 'max_buffer_len_factor': 6.0, 'max_seq_len': 4608, 'n_head': 4, 'n_mha_win_size': 19, 'num_classes': 32, 'regression_range': [(0, 4), (4, 8), (8, 16), (16, 32), (32, 64), (64, 10000)], 'scale_factor': 2, 'test_cfg': {'duration_thresh': 0.05, 'ext_score_file': None, 'iou_threshold': 0.1, 'max_seg_num': 200, 'min_score': 0.001, 'multiclass_nms': True, 'nms_method': 'soft', 'nms_sigma': 0.5, 'pre_nms_thresh': 0.001, 'pre_nms_topk': 2000, 'voting_thresh': 0.7}, 'train_cfg': {'center_sample': 'radius', 'center_sample_radius': 1.5, 'clip_grad_l2norm': 1.0, 'cls_prior_prob': 0.01, 'dropout': 0.0, 'droppath': 0.1, 'head_empty_cls': [], 'init_loss_norm': 100, 'label_smoothing': 0.0, 'loss_weight': 1.0}, 'use_abs_pe': False, 'use_rel_pe': False}, 'model_name': 'ScoreTransformer', 'opt': {'epochs': 60, 'learning_rate': 0.0001, 'momentum': 0.9, 'schedule_gamma': 0.1, 'schedule_steps': [], 'schedule_type': 'cosine', 'type': 'AdamW', 'warmup': True, 'warmup_epochs': 5, 'weight_decay': 0.05}, 'output_folder': './ckpt/', 'score_model': {'attn_ratio': 4, 'hidden_size': 64, 'input_size': 512, 'score_size': [2, 6, 6]}, 'test_cfg': {'duration_thresh': 0.05, 'ext_score_file': None, 'iou_threshold': 0.1, 'max_seg_num': 200, 'min_score': 0.001, 'multiclass_nms': True, 'nms_method': 'soft', 'nms_sigma': 0.5, 'pre_nms_thresh': 0.001, 'pre_nms_topk': 2000, 'voting_thresh': 0.7}, 'train_cfg': {'center_sample': 'radius', 'center_sample_radius': 1.5, 'clip_grad_l2norm': 1.0, 'cls_prior_prob': 0.01, 'dropout': 0.0, 'droppath': 0.1, 'head_empty_cls': [], 'init_loss_norm': 100, 'label_smoothing': 0.0, 'loss_weight': 1.0}, 'train_split': ['validation'], 'val_split': ['validation']} grade = 1 , count = 219 grade = 0 , count = 23 => loading checkpoint './ckpt/74_resnet50_score74A2/epoch_064.pth.tar' Loading from EMA model ... Start testing model ScoreTransformer ... Test: [00010/00242] Time 0.40 (0.40) Test: [00020/00242] Time 0.13 (0.26) Test: [00030/00242] Time 0.11 (0.21) Test: [00040/00242] Time 0.12 (0.19) Test: [00050/00242] Time 0.12 (0.17) Test: [00060/00242] Time 0.12 (0.17) Test: [00070/00242] Time 0.12 (0.16) Test: [00080/00242] Time 0.12 (0.16) Test: [00090/00242] Time 0.12 (0.15) Test: [00100/00242] Time 0.12 (0.15) Test: [00110/00242] Time 0.12 (0.15) Test: [00120/00242] Time 0.12 (0.14) Test: [00130/00242] Time 0.12 (0.14) Test: [00140/00242] Time 0.12 (0.14) Test: [00150/00242] Time 0.12 (0.14) Test: [00160/00242] Time 0.12 (0.14) Test: [00170/00242] Time 0.12 (0.14) Test: [00180/00242] Time 0.12 (0.14) Test: [00190/00242] Time 0.12 (0.14) Test: [00200/00242] Time 0.11 (0.13) Test: [00210/00242] Time 0.11 (0.13) Test: [00220/00242] Time 0.12 (0.13) Test: [00230/00242] Time 0.12 (0.13) Test: [00240/00242] Time 0.12 (0.13) [RESULTS] Action detection results on 74_anno_score_4myres2net. |tIoU = 0.30: mAP = 93.28 (%) |tIoU = 0.40: mAP = 93.12 (%) |tIoU = 0.50: mAP = 91.10 (%) |tIoU = 0.60: mAP = 90.12 (%) |tIoU = 0.70: mAP = 83.16 (%) Avearge mAP: 90.15 (%) grade: precision recall f1-score support 0 0.14 1.00 0.25 23 1 1.00 0.37 0.54 219 accuracy 0.43 242 macro avg 0.57 0.68 0.39 242 weighted avg 0.92 0.43 0.51 242 scores 1: precision recall f1-score support 0 0.41 0.99 0.58 84 1 0.58 0.19 0.29 57 2 0.00 0.00 0.00 1 3 0.00 0.00 0.00 17 4 0.00 0.00 0.00 5 5 0.90 0.23 0.37 78 accuracy 0.46 242 macro avg 0.31 0.24 0.21 242 weighted avg 0.57 0.46 0.39 242 scores 2: precision recall f1-score support 1 0.00 0.00 0.00 11 2 0.00 0.00 0.00 3 3 0.00 0.00 0.00 7 4 0.47 0.28 0.35 25 5 0.83 0.96 0.89 196 accuracy 0.81 242 macro avg 0.26 0.25 0.25 242 weighted avg 0.72 0.81 0.76 242 All done! Total time: 80.91 sec