测试结果:

#PAN Full

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
# python test_models.py something --VAP --batch_size=20 -j=4 --test_crops=1 --test_segments=8,8 --weights=pretrained/PAN_RGB_something_resnet50_shift8_blockres_avg_segment8_e50.pth.tar,pretrained/PAN_PA_something_resnet50_shift8_blockres_avg_segment8_e80.pth.tar
$ bash scripts/test/sthv1/Full.sh
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: RGB
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
video number:11522
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: PA
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
video number:11522
video 0 done, total 0/11522, average 0.229 sec/video, moving Prec@1 50.000 Prec@5 85.000
video 400 done, total 400/11522, average 0.032 sec/video, moving Prec@1 52.143 Prec@5 83.333
video 800 done, total 800/11522, average 0.028 sec/video, moving Prec@1 51.341 Prec@5 80.732
video 1200 done, total 1200/11522, average 0.026 sec/video, moving Prec@1 51.557 Prec@5 80.738
video 1600 done, total 1600/11522, average 0.025 sec/video, moving Prec@1 51.728 Prec@5 80.432
video 2000 done, total 2000/11522, average 0.025 sec/video, moving Prec@1 50.792 Prec@5 79.950
video 2400 done, total 2400/11522, average 0.024 sec/video, moving Prec@1 50.744 Prec@5 79.628
video 2800 done, total 2800/11522, average 0.024 sec/video, moving Prec@1 50.887 Prec@5 79.716
video 3200 done, total 3200/11522, average 0.024 sec/video, moving Prec@1 50.559 Prec@5 79.720
video 3600 done, total 3600/11522, average 0.024 sec/video, moving Prec@1 50.497 Prec@5 79.586
video 4000 done, total 4000/11522, average 0.023 sec/video, moving Prec@1 50.274 Prec@5 79.254
video 4400 done, total 4400/11522, average 0.023 sec/video, moving Prec@1 50.204 Prec@5 79.072
video 4800 done, total 4800/11522, average 0.023 sec/video, moving Prec@1 50.083 Prec@5 78.963
video 5200 done, total 5200/11522, average 0.023 sec/video, moving Prec@1 50.057 Prec@5 78.812
video 5600 done, total 5600/11522, average 0.023 sec/video, moving Prec@1 50.356 Prec@5 78.968
video 6000 done, total 6000/11522, average 0.023 sec/video, moving Prec@1 50.233 Prec@5 78.887
video 6400 done, total 6400/11522, average 0.023 sec/video, moving Prec@1 50.327 Prec@5 79.065
video 6800 done, total 6800/11522, average 0.023 sec/video, moving Prec@1 50.249 Prec@5 78.974
video 7200 done, total 7200/11522, average 0.023 sec/video, moving Prec@1 50.208 Prec@5 78.934
video 7600 done, total 7600/11522, average 0.023 sec/video, moving Prec@1 50.367 Prec@5 78.950
video 8000 done, total 8000/11522, average 0.023 sec/video, moving Prec@1 50.274 Prec@5 79.077
video 8400 done, total 8400/11522, average 0.023 sec/video, moving Prec@1 50.226 Prec@5 79.062
video 8800 done, total 8800/11522, average 0.023 sec/video, moving Prec@1 50.227 Prec@5 79.138
video 9200 done, total 9200/11522, average 0.023 sec/video, moving Prec@1 50.195 Prec@5 79.143
video 9600 done, total 9600/11522, average 0.023 sec/video, moving Prec@1 50.364 Prec@5 79.179
video 10000 done, total 10000/11522, average 0.023 sec/video, moving Prec@1 50.379 Prec@5 79.232
video 10400 done, total 10400/11522, average 0.023 sec/video, moving Prec@1 50.480 Prec@5 79.290
video 10800 done, total 10800/11522, average 0.023 sec/video, moving Prec@1 50.582 Prec@5 79.307
video 11200 done, total 11200/11522, average 0.023 sec/video, moving Prec@1 50.606 Prec@5 79.269
[0.89552239 0.3125 0.26666667 0.65671642 0.36842105 0.48275862
0.70895522 0.38235294 0.62135922 0.70873786 0.44 0.40277778
0.30851064 0.35483871 0.77070064 0.4 0.24096386 0.2625
0.45528455 0.34666667 0.352 0.60344828 0.46296296 0.45098039
0.51612903 0.41666667 0.43478261 0.4 0.6969697 0.61904762
0.59090909 0.53846154 0.87037037 0.25274725 0.23913043 0.13888889
0.7816092 0.79411765 0.07894737 0.59259259 0.68595041 0.64
0.67692308 0.67346939 0.75 0.57377049 0.47794118 0.36842105
0.15730337 0.68421053 0.61320755 0. 0.36 0.45
0.19047619 0.25 0.45454545 0.5 0. 0.74489796
0.75757576 0.13888889 0.64 0.1 0.1 0.46296296
0.18518519 0.21212121 0.43661972 0.34615385 0.5483871 0.21428571
0.55263158 0.48 0.83333333 0.41935484 0.35294118 0.58823529
0.27272727 0.82926829 0.11428571 0.21428571 0.33333333 0.15789474
0.6 0.39655172 0.60810811 0.6984127 0.27777778 0.07142857
0.59459459 0.75609756 0.58139535 0.55555556 0.62616822 0.22580645
0.03703704 0.11904762 0.57894737 0.45588235 0.32653061 0.51851852
0.203125 0.53061224 0.68292683 0.63809524 0.52808989 0.73913043
0.49122807 0.32967033 0.58823529 0. 0.66216216 0.28571429
0.60377358 0.43333333 0.27272727 0.73404255 0.29268293 0.54166667
0.80681818 0.60784314 0.35454545 0.5 0.33333333 0.44897959
0.47169811 0.57407407 0.14482759 0.63461538 0.22 0.22222222
0.23684211 0.36666667 0.68874172 0.44827586 0. 0.31666667
0.13559322 0.56578947 0.66 0.39534884 0.45454545 0.43220339
0.6 0.47183099 0.63207547 0.17910448 0.546875 0.8
0.66666667 0.29032258 0.38333333 0.7962963 0.55952381 0.11764706
0.29508197 0.3 0.47524752 0.34375 0.42592593 0.325
0.3 0.37931034 0.69607843 0.77272727 0.9223301 0.87735849
0.74074074 0.75675676 0.45714286 0.62589928 0.69343066 0.69736842]
upper bound: 0.483756929950145
-----Evaluation is finished------
Class Accuracy 46.26%
Overall Prec@1 50.53% Prec@5 79.23%
/home/hs/anaconda3/envs/dzl/lib/python3.7/site-packages/numpy/core/_asarray.py:136: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
return array(a, dtype, copy=False, order=order, subok=True)

#PAN Full

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
# python test_models.py something --VAP --batch_size=20 -j=4 --test_crops=3 --test_segments=8,8 --weights=pretrained/PAN_RGB_something_resnet50_shift8_blockres_avg_segment8_e50.pth.tar,pretrained/PAN_PA_something_resnet50_shift8_blockres_avg_segment8_e80.pth.tar 
$ bash scripts/test/sthv1/Full.sh
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: RGB
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
video number:11522
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: PA
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
video number:11522
video 0 done, total 0/11522, average 0.469 sec/video, moving Prec@1 55.000 Prec@5 85.000
video 400 done, total 400/11522, average 0.083 sec/video, moving Prec@1 54.524 Prec@5 83.810
video 800 done, total 800/11522, average 0.074 sec/video, moving Prec@1 52.317 Prec@5 80.366
video 1200 done, total 1200/11522, average 0.071 sec/video, moving Prec@1 52.377 Prec@5 80.902
video 1600 done, total 1600/11522, average 0.069 sec/video, moving Prec@1 52.469 Prec@5 80.494
video 2000 done, total 2000/11522, average 0.069 sec/video, moving Prec@1 51.634 Prec@5 80.198
video 2400 done, total 2400/11522, average 0.068 sec/video, moving Prec@1 51.488 Prec@5 80.248
video 2800 done, total 2800/11522, average 0.068 sec/video, moving Prec@1 51.596 Prec@5 80.319
video 3200 done, total 3200/11522, average 0.067 sec/video, moving Prec@1 51.273 Prec@5 80.217
video 3600 done, total 3600/11522, average 0.067 sec/video, moving Prec@1 51.077 Prec@5 80.193
video 4000 done, total 4000/11522, average 0.067 sec/video, moving Prec@1 50.821 Prec@5 79.801
video 4400 done, total 4400/11522, average 0.067 sec/video, moving Prec@1 50.814 Prec@5 79.570
video 4800 done, total 4800/11522, average 0.067 sec/video, moving Prec@1 50.705 Prec@5 79.378
video 5200 done, total 5200/11522, average 0.067 sec/video, moving Prec@1 50.632 Prec@5 79.253
video 5600 done, total 5600/11522, average 0.067 sec/video, moving Prec@1 50.925 Prec@5 79.253
video 6000 done, total 6000/11522, average 0.066 sec/video, moving Prec@1 50.831 Prec@5 79.236
video 6400 done, total 6400/11522, average 0.066 sec/video, moving Prec@1 50.857 Prec@5 79.377
video 6800 done, total 6800/11522, average 0.066 sec/video, moving Prec@1 50.850 Prec@5 79.355
video 7200 done, total 7200/11522, average 0.066 sec/video, moving Prec@1 50.748 Prec@5 79.224
video 7600 done, total 7600/11522, average 0.066 sec/video, moving Prec@1 50.853 Prec@5 79.265
video 8000 done, total 8000/11522, average 0.066 sec/video, moving Prec@1 50.736 Prec@5 79.389
video 8400 done, total 8400/11522, average 0.066 sec/video, moving Prec@1 50.760 Prec@5 79.394
video 8800 done, total 8800/11522, average 0.066 sec/video, moving Prec@1 50.873 Prec@5 79.433
video 9200 done, total 9200/11522, average 0.066 sec/video, moving Prec@1 50.824 Prec@5 79.425
video 9600 done, total 9600/11522, average 0.066 sec/video, moving Prec@1 50.977 Prec@5 79.407
video 10000 done, total 10000/11522, average 0.066 sec/video, moving Prec@1 50.948 Prec@5 79.451
video 10400 done, total 10400/11522, average 0.066 sec/video, moving Prec@1 51.046 Prec@5 79.472
video 10800 done, total 10800/11522, average 0.066 sec/video, moving Prec@1 51.063 Prec@5 79.547
video 11200 done, total 11200/11522, average 0.066 sec/video, moving Prec@1 51.168 Prec@5 79.590
[0.92537313 0.375 0.26666667 0.62686567 0.39473684 0.45689655
0.67910448 0.44117647 0.59223301 0.70873786 0.48 0.40277778
0.29787234 0.35483871 0.75159236 0.44 0.24698795 0.275
0.47154472 0.37333333 0.352 0.56896552 0.57407407 0.45098039
0.51612903 0.5 0.43478261 0.44 0.66666667 0.68253968
0.63636364 0.53846154 0.87037037 0.18681319 0.26086957 0.13888889
0.77011494 0.80392157 0.05263158 0.55555556 0.71900826 0.64
0.71538462 0.69387755 0.76923077 0.60655738 0.5 0.36842105
0.20224719 0.69924812 0.61320755 0. 0.4 0.45
0.19047619 0.20833333 0.48484848 0.5 0. 0.81632653
0.78787879 0.16666667 0.64 0.06363636 0.1 0.40740741
0.16666667 0.21212121 0.46478873 0.30769231 0.5483871 0.25
0.55263158 0.48 0.83333333 0.4516129 0.29411765 0.58823529
0.22727273 0.80487805 0.17142857 0.26190476 0.36666667 0.15789474
0.6 0.4137931 0.63513514 0.73015873 0.16666667 0.05357143
0.59459459 0.73170732 0.46511628 0.55555556 0.6728972 0.29032258
0.03703704 0.0952381 0.56725146 0.48529412 0.44897959 0.5462963
0.25 0.53061224 0.67073171 0.67619048 0.51685393 0.72463768
0.40350877 0.31868132 0.60784314 0. 0.66216216 0.28571429
0.62264151 0.4 0.36363636 0.74468085 0.26829268 0.5625
0.73863636 0.58823529 0.35454545 0.475 0.30555556 0.46938776
0.39622642 0.62962963 0.13793103 0.64423077 0.22 0.18518519
0.21052632 0.33333333 0.70198675 0.44827586 0. 0.31666667
0.13559322 0.56578947 0.66 0.44186047 0.45454545 0.44915254
0.65333333 0.47887324 0.59433962 0.19402985 0.578125 0.816
0.64848485 0.29032258 0.38333333 0.81481481 0.51190476 0.17647059
0.29508197 0.31428571 0.46534653 0.375 0.46296296 0.3
0.34 0.31034483 0.69607843 0.8030303 0.9223301 0.88679245
0.71296296 0.7027027 0.44761905 0.66906475 0.71532847 0.72368421]
upper bound: 0.48772172111390605
-----Evaluation is finished------
Class Accuracy 46.72%
Overall Prec@1 51.10% Prec@5 79.56%
/home/hs/anaconda3/envs/dzl/lib/python3.7/site-packages/numpy/core/_asarray.py:136: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
return array(a, dtype, copy=False, order=order, subok=True)

#PAN En

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
# python test_models.py something \
# --VAP --batch_size=16 -j=4 --test_crops=1 --test_segments=8,8,8 \
# --weights=pretrained/PAN_Lite_something_resnet50_shift8_blockres_avg_segment8_e80.pth.tar,pretrained/PAN_RGB_something_resnet50_shift8_blockres_avg_segment8_e50.pth.tar,pretrained/PAN_PA_something_resnet50_shift8_blockres_avg_segment8_e80.pth.tar \
# --full_res --twice_sample
$ bash scripts/test/sthv1/En.sh
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: Lite
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
=> Converting the ImageNet model to a PAN_Lite init model
=> Done. PAN_lite model ready...
=> Using twice sample for the dataset...
video number:11522
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: RGB
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
=> Using twice sample for the dataset...
video number:11522
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: PA
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
=> Using twice sample for the dataset...
video number:11522
video 0 done, total 0/11522, average 0.991 sec/video, moving Prec@1 50.000 Prec@5 75.000
video 320 done, total 320/11522, average 0.151 sec/video, moving Prec@1 55.655 Prec@5 85.714
video 640 done, total 640/11522, average 0.148 sec/video, moving Prec@1 55.488 Prec@5 83.689
video 960 done, total 960/11522, average 0.146 sec/video, moving Prec@1 54.303 Prec@5 82.377
video 1280 done, total 1280/11522, average 0.143 sec/video, moving Prec@1 55.710 Prec@5 83.025
video 1600 done, total 1600/11522, average 0.142 sec/video, moving Prec@1 55.941 Prec@5 82.735
video 1920 done, total 1920/11522, average 0.141 sec/video, moving Prec@1 55.269 Prec@5 82.386
video 2240 done, total 2240/11522, average 0.141 sec/video, moving Prec@1 55.186 Prec@5 82.225
video 2560 done, total 2560/11522, average 0.142 sec/video, moving Prec@1 54.891 Prec@5 81.793
video 2880 done, total 2880/11522, average 0.141 sec/video, moving Prec@1 54.454 Prec@5 81.872
video 3200 done, total 3200/11522, average 0.141 sec/video, moving Prec@1 53.887 Prec@5 81.965
video 3520 done, total 3520/11522, average 0.141 sec/video, moving Prec@1 53.790 Prec@5 81.900
video 3840 done, total 3840/11522, average 0.141 sec/video, moving Prec@1 53.994 Prec@5 81.613
video 4160 done, total 4160/11522, average 0.141 sec/video, moving Prec@1 53.855 Prec@5 81.418
video 4480 done, total 4480/11522, average 0.140 sec/video, moving Prec@1 53.625 Prec@5 81.406
video 4800 done, total 4800/11522, average 0.140 sec/video, moving Prec@1 53.488 Prec@5 81.146
video 5120 done, total 5120/11522, average 0.140 sec/video, moving Prec@1 53.154 Prec@5 80.880
video 5440 done, total 5440/11522, average 0.140 sec/video, moving Prec@1 53.372 Prec@5 81.067
video 5760 done, total 5760/11522, average 0.140 sec/video, moving Prec@1 53.307 Prec@5 80.956
video 6080 done, total 6080/11522, average 0.140 sec/video, moving Prec@1 53.264 Prec@5 81.053
video 6400 done, total 6400/11522, average 0.140 sec/video, moving Prec@1 53.304 Prec@5 81.188
video 6720 done, total 6720/11522, average 0.140 sec/video, moving Prec@1 53.192 Prec@5 81.205
video 7040 done, total 7040/11522, average 0.140 sec/video, moving Prec@1 53.104 Prec@5 81.151
video 7360 done, total 7360/11522, average 0.140 sec/video, moving Prec@1 53.037 Prec@5 81.047
video 7680 done, total 7680/11522, average 0.140 sec/video, moving Prec@1 53.015 Prec@5 81.120
video 8000 done, total 8000/11522, average 0.140 sec/video, moving Prec@1 53.031 Prec@5 81.125
video 8320 done, total 8320/11522, average 0.140 sec/video, moving Prec@1 53.071 Prec@5 81.214
video 8640 done, total 8640/11522, average 0.140 sec/video, moving Prec@1 52.830 Prec@5 81.134
video 8960 done, total 8960/11522, average 0.140 sec/video, moving Prec@1 52.852 Prec@5 81.116
video 9280 done, total 9280/11522, average 0.140 sec/video, moving Prec@1 52.829 Prec@5 81.142
video 9600 done, total 9600/11522, average 0.140 sec/video, moving Prec@1 52.943 Prec@5 81.136
video 9920 done, total 9920/11522, average 0.140 sec/video, moving Prec@1 52.909 Prec@5 81.119
video 10240 done, total 10240/11522, average 0.140 sec/video, moving Prec@1 53.013 Prec@5 81.260
video 10560 done, total 10560/11522, average 0.140 sec/video, moving Prec@1 53.064 Prec@5 81.203
video 10880 done, total 10880/11522, average 0.140 sec/video, moving Prec@1 53.111 Prec@5 81.232
video 11200 done, total 11200/11522, average 0.140 sec/video, moving Prec@1 53.174 Prec@5 81.232
video 11520 done, total 11520/11522, average 0.139 sec/video, moving Prec@1 53.203 Prec@5 81.236
[0.91044776 0.4125 0.31428571 0.65671642 0.5 0.50862069
0.69402985 0.47058824 0.72815534 0.75728155 0.50666667 0.54166667
0.41489362 0.41935484 0.80254777 0.48 0.24096386 0.325
0.4796748 0.38666667 0.408 0.63793103 0.5 0.43137255
0.48387097 0.41666667 0.56521739 0.32 0.72727273 0.73015873
0.63636364 0.53846154 0.85185185 0.25274725 0.26086957 0.22222222
0.7816092 0.85294118 0.07894737 0.62962963 0.71900826 0.64
0.70769231 0.71428571 0.71153846 0.67213115 0.47058824 0.41052632
0.28089888 0.70676692 0.60377358 0. 0.36 0.55
0.23809524 0.16666667 0.51515152 0.4375 0. 0.7755102
0.81818182 0.25 0.6 0.1 0.16666667 0.5462963
0.09259259 0.21212121 0.46478873 0.42307692 0.51612903 0.25
0.52631579 0.52 0.83333333 0.58064516 0.23529412 0.58823529
0.36363636 0.82926829 0.11428571 0.14285714 0.4 0.15789474
0.57142857 0.4137931 0.63513514 0.79365079 0.16666667 0.08928571
0.51351351 0.70731707 0.53488372 0.62037037 0.73831776 0.27419355
0.03703704 0.19047619 0.59064327 0.5 0.3877551 0.58333333
0.25 0.46938776 0.70731707 0.7047619 0.56179775 0.71014493
0.49122807 0.36263736 0.64705882 0. 0.64864865 0.38095238
0.66981132 0.46666667 0.32727273 0.75531915 0.34146341 0.625
0.81818182 0.60784314 0.39090909 0.45 0.27777778 0.48979592
0.47169811 0.62962963 0.16551724 0.70192308 0.3 0.22222222
0.34210526 0.26666667 0.70198675 0.48275862 0.14285714 0.2
0.13559322 0.64473684 0.67 0.27906977 0.54545455 0.48305085
0.54666667 0.5 0.62264151 0.20895522 0.515625 0.744
0.69090909 0.25806452 0.36666667 0.87037037 0.5952381 0.21568627
0.24590164 0.37142857 0.5049505 0.34375 0.48148148 0.35
0.34 0.34482759 0.70588235 0.81818182 0.9223301 0.93396226
0.74074074 0.72972973 0.44761905 0.64028777 0.71532847 0.72368421]
upper bound: 0.5058494457198016
-----Evaluation is finished------
Class Accuracy 48.66%
Overall Prec@1 53.20% Prec@5 81.24%
/home/hs/anaconda3/envs/dzl/lib/python3.7/site-packages/numpy/core/_asarray.py:136: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
return array(a, dtype, copy=False, order=order, subok=True)

#PAN PA

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
$ python test_models.py something --VAP --batch_size=10 -j=4 --test_crops=1 --test_segments=8 --weights=pretrained/PAN_PA_something_resnet50_shift8_blockres_avg_segment8_e80.pth.tar
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: PA
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
video number:11522
video 0 done, total 0/11522, average 13.141 sec/video, moving Prec@1 50.000 Prec@5 80.000
video 200 done, total 200/11522, average 0.950 sec/video, moving Prec@1 48.095 Prec@5 82.857
video 400 done, total 400/11522, average 0.675 sec/video, moving Prec@1 47.805 Prec@5 80.244
video 600 done, total 600/11522, average 0.586 sec/video, moving Prec@1 49.836 Prec@5 80.164
video 800 done, total 800/11522, average 0.537 sec/video, moving Prec@1 47.037 Prec@5 78.148
video 1000 done, total 1000/11522, average 0.480 sec/video, moving Prec@1 46.139 Prec@5 77.624
video 1200 done, total 1200/11522, average 0.449 sec/video, moving Prec@1 47.273 Prec@5 77.686
video 1400 done, total 1400/11522, average 0.440 sec/video, moving Prec@1 47.163 Prec@5 77.021
video 1600 done, total 1600/11522, average 0.420 sec/video, moving Prec@1 46.957 Prec@5 76.522
video 1800 done, total 1800/11522, average 0.406 sec/video, moving Prec@1 46.188 Prec@5 76.133
video 2000 done, total 2000/11522, average 0.387 sec/video, moving Prec@1 46.070 Prec@5 76.070
video 2200 done, total 2200/11522, average 0.368 sec/video, moving Prec@1 46.018 Prec@5 75.928
video 2400 done, total 2400/11522, average 0.358 sec/video, moving Prec@1 46.307 Prec@5 75.187
video 2600 done, total 2600/11522, average 0.350 sec/video, moving Prec@1 46.398 Prec@5 75.364
video 2800 done, total 2800/11522, average 0.336 sec/video, moving Prec@1 46.619 Prec@5 75.480
video 3000 done, total 3000/11522, average 0.330 sec/video, moving Prec@1 46.611 Prec@5 75.714
video 3200 done, total 3200/11522, average 0.325 sec/video, moving Prec@1 46.386 Prec@5 75.607
video 3400 done, total 3400/11522, average 0.318 sec/video, moving Prec@1 46.305 Prec@5 75.630
video 3600 done, total 3600/11522, average 0.311 sec/video, moving Prec@1 46.260 Prec@5 75.208
video 3800 done, total 3800/11522, average 0.304 sec/video, moving Prec@1 46.194 Prec@5 75.144
video 4000 done, total 4000/11522, average 0.297 sec/video, moving Prec@1 46.010 Prec@5 74.988
video 4200 done, total 4200/11522, average 0.293 sec/video, moving Prec@1 45.914 Prec@5 74.893
video 4400 done, total 4400/11522, average 0.288 sec/video, moving Prec@1 45.782 Prec@5 74.807
video 4600 done, total 4600/11522, average 0.290 sec/video, moving Prec@1 45.792 Prec@5 74.555
video 4800 done, total 4800/11522, average 0.289 sec/video, moving Prec@1 45.530 Prec@5 74.470
video 5000 done, total 5000/11522, average 0.285 sec/video, moving Prec@1 45.389 Prec@5 74.311
video 5200 done, total 5200/11522, average 0.283 sec/video, moving Prec@1 45.528 Prec@5 74.299
video 5400 done, total 5400/11522, average 0.282 sec/video, moving Prec@1 45.564 Prec@5 74.251
video 5600 done, total 5600/11522, average 0.279 sec/video, moving Prec@1 45.704 Prec@5 74.314
video 5800 done, total 5800/11522, average 0.276 sec/video, moving Prec@1 45.611 Prec@5 74.148
video 6000 done, total 6000/11522, average 0.273 sec/video, moving Prec@1 45.674 Prec@5 74.260
video 6200 done, total 6200/11522, average 0.270 sec/video, moving Prec@1 45.749 Prec@5 74.235
video 6400 done, total 6400/11522, average 0.266 sec/video, moving Prec@1 45.803 Prec@5 74.368
video 6600 done, total 6600/11522, average 0.264 sec/video, moving Prec@1 45.719 Prec@5 74.433
video 6800 done, total 6800/11522, average 0.261 sec/video, moving Prec@1 45.551 Prec@5 74.347
video 7000 done, total 7000/11522, average 0.258 sec/video, moving Prec@1 45.578 Prec@5 74.508
video 7200 done, total 7200/11522, average 0.255 sec/video, moving Prec@1 45.423 Prec@5 74.452
video 7400 done, total 7400/11522, average 0.252 sec/video, moving Prec@1 45.452 Prec@5 74.345
video 7600 done, total 7600/11522, average 0.250 sec/video, moving Prec@1 45.427 Prec@5 74.428
video 7800 done, total 7800/11522, average 0.249 sec/video, moving Prec@1 45.314 Prec@5 74.507
video 8000 done, total 8000/11522, average 0.246 sec/video, moving Prec@1 45.268 Prec@5 74.469
video 8200 done, total 8200/11522, average 0.243 sec/video, moving Prec@1 45.311 Prec@5 74.531
video 8400 done, total 8400/11522, average 0.241 sec/video, moving Prec@1 45.268 Prec@5 74.530
video 8600 done, total 8600/11522, average 0.239 sec/video, moving Prec@1 45.203 Prec@5 74.553
video 8800 done, total 8800/11522, average 0.237 sec/video, moving Prec@1 45.255 Prec@5 74.620
video 9000 done, total 9000/11522, average 0.236 sec/video, moving Prec@1 45.161 Prec@5 74.584
video 9200 done, total 9200/11522, average 0.234 sec/video, moving Prec@1 45.179 Prec@5 74.604
video 9400 done, total 9400/11522, average 0.232 sec/video, moving Prec@1 45.228 Prec@5 74.633
video 9600 done, total 9600/11522, average 0.230 sec/video, moving Prec@1 45.265 Prec@5 74.631
video 9800 done, total 9800/11522, average 0.228 sec/video, moving Prec@1 45.301 Prec@5 74.699
video 10000 done, total 10000/11522, average 0.227 sec/video, moving Prec@1 45.305 Prec@5 74.715
video 10200 done, total 10200/11522, average 0.225 sec/video, moving Prec@1 45.260 Prec@5 74.760
video 10400 done, total 10400/11522, average 0.224 sec/video, moving Prec@1 45.312 Prec@5 74.774
video 10600 done, total 10600/11522, average 0.223 sec/video, moving Prec@1 45.297 Prec@5 74.807
video 10800 done, total 10800/11522, average 0.225 sec/video, moving Prec@1 45.412 Prec@5 74.875
video 11000 done, total 11000/11522, average 0.223 sec/video, moving Prec@1 45.441 Prec@5 74.877
video 11200 done, total 11200/11522, average 0.222 sec/video, moving Prec@1 45.468 Prec@5 74.844
video 11400 done, total 11400/11522, average 0.220 sec/video, moving Prec@1 45.451 Prec@5 74.812
[0.86567164 0.25 0.24761905 0.59701493 0.39473684 0.45689655
0.62686567 0.29411765 0.52427184 0.62135922 0.4 0.41666667
0.25531915 0.32258065 0.61146497 0.42 0.23493976 0.25
0.34146341 0.28 0.248 0.60344828 0.5 0.35294118
0.5483871 0.33333333 0.43478261 0.4 0.66666667 0.61904762
0.54545455 0.5 0.75925926 0.15384615 0.26086957 0.19444444
0.81609195 0.7745098 0.13157895 0.55555556 0.69421488 0.66
0.60769231 0.55102041 0.71153846 0.60655738 0.45588235 0.33684211
0.21348315 0.59398496 0.48113208 0. 0.28 0.45
0.19047619 0.16666667 0.36363636 0.46875 0.08333333 0.70408163
0.81818182 0.13888889 0.56 0.13636364 0.16666667 0.34259259
0.12962963 0.10606061 0.35211268 0.34615385 0.48387097 0.17857143
0.39473684 0.48 0.75 0.32258065 0.23529412 0.47058824
0.40909091 0.80487805 0.08571429 0.21428571 0.36666667 0.15789474
0.6 0.34482759 0.63513514 0.63492063 0.05555556 0.03571429
0.56756757 0.70731707 0.48837209 0.50925926 0.61682243 0.16129032
0. 0.16666667 0.58479532 0.48529412 0.41836735 0.42592593
0.15625 0.57142857 0.6097561 0.54285714 0.40449438 0.56521739
0.31578947 0.34065934 0.58823529 0. 0.5 0.19047619
0.38679245 0.43333333 0.25454545 0.61702128 0.26829268 0.54166667
0.71590909 0.58823529 0.30909091 0.425 0.19444444 0.28571429
0.43396226 0.46296296 0.11034483 0.60576923 0.24 0.11111111
0.26315789 0.33333333 0.52317881 0.39655172 0. 0.11666667
0.10169492 0.60526316 0.61 0.23255814 0.38636364 0.33050847
0.49333333 0.45070423 0.52830189 0.14925373 0.4375 0.68
0.65454545 0.33870968 0.4 0.77777778 0.48809524 0.07843137
0.31147541 0.35714286 0.35643564 0.1875 0.48148148 0.275
0.3 0.31034483 0.64705882 0.75757576 0.90291262 0.8490566
0.69444444 0.72972973 0.31428571 0.56834532 0.62773723 0.55263158]
upper bound: 0.4377580829893891
-----Evaluation is finished------
Class Accuracy 41.77%
Overall Prec@1 45.42% Prec@5 74.83%
E:\Program Files\Anaconda3\envs\tsm\lib\site-packages\numpy\core\_asarray.py:136: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray return array(a, dtype, copy=False, order=order, subok=True)

#PAN RGB

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
$ python test_models.py something --VAP --batch_size=10 -j=4 --test_crops=1 --test_segments=8 --weights=pretrained/PAN_RGB_something_resnet50_shift8_blockres_avg_segment8_e50.pth.tar
something: 174 classes
=> shift: True, shift_div: 8, shift_place: blockres

Initializing PAN with base model: resnet50.
PAN Configurations:
input_modality: RGB
num_segments: 8
new_length: 1
consensus_module: avg
dropout_ratio: 0.8
img_feature_dim: 256

=> base model: resnet50
Adding temporal shift...
=> Using 3-level VAP
video number:11522
video 0 done, total 0/11522, average 3.020 sec/video, moving Prec@1 30.000 Prec@5 70.000
video 200 done, total 200/11522, average 0.238 sec/video, moving Prec@1 48.571 Prec@5 77.143
video 400 done, total 400/11522, average 0.174 sec/video, moving Prec@1 47.805 Prec@5 77.073
video 600 done, total 600/11522, average 0.146 sec/video, moving Prec@1 47.377 Prec@5 77.213
video 800 done, total 800/11522, average 0.134 sec/video, moving Prec@1 45.556 Prec@5 75.309
video 1000 done, total 1000/11522, average 0.136 sec/video, moving Prec@1 46.139 Prec@5 74.554
video 1200 done, total 1200/11522, average 0.132 sec/video, moving Prec@1 45.785 Prec@5 75.124
video 1400 done, total 1400/11522, average 0.126 sec/video, moving Prec@1 46.596 Prec@5 74.894
video 1600 done, total 1600/11522, average 0.121 sec/video, moving Prec@1 46.770 Prec@5 74.845
video 1800 done, total 1800/11522, average 0.121 sec/video, moving Prec@1 45.691 Prec@5 74.088
video 2000 done, total 2000/11522, average 0.123 sec/video, moving Prec@1 45.821 Prec@5 74.577
video 2200 done, total 2200/11522, average 0.122 sec/video, moving Prec@1 45.520 Prec@5 74.253
video 2400 done, total 2400/11522, average 0.122 sec/video, moving Prec@1 45.270 Prec@5 74.191
video 2600 done, total 2600/11522, average 0.120 sec/video, moving Prec@1 45.057 Prec@5 74.215
video 2800 done, total 2800/11522, average 0.117 sec/video, moving Prec@1 45.089 Prec@5 74.377
video 3000 done, total 3000/11522, average 0.114 sec/video, moving Prec@1 45.083 Prec@5 74.319
video 3200 done, total 3200/11522, average 0.112 sec/video, moving Prec@1 44.517 Prec@5 74.050
video 3400 done, total 3400/11522, average 0.110 sec/video, moving Prec@1 44.516 Prec@5 74.164
video 3600 done, total 3600/11522, average 0.109 sec/video, moving Prec@1 44.515 Prec@5 74.072
video 3800 done, total 3800/11522, average 0.107 sec/video, moving Prec@1 44.514 Prec@5 73.858
video 4000 done, total 4000/11522, average 0.106 sec/video, moving Prec@1 44.214 Prec@5 73.791
video 4200 done, total 4200/11522, average 0.106 sec/video, moving Prec@1 44.228 Prec@5 73.753
video 4400 done, total 4400/11522, average 0.103 sec/video, moving Prec@1 44.331 Prec@5 73.651
video 4600 done, total 4600/11522, average 0.104 sec/video, moving Prec@1 44.555 Prec@5 73.557
video 4800 done, total 4800/11522, average 0.102 sec/video, moving Prec@1 44.470 Prec@5 73.451
video 5000 done, total 5000/11522, average 0.102 sec/video, moving Prec@1 44.311 Prec@5 73.234
video 5200 done, total 5200/11522, average 0.101 sec/video, moving Prec@1 44.376 Prec@5 73.378
video 5400 done, total 5400/11522, average 0.100 sec/video, moving Prec@1 44.455 Prec@5 73.420
video 5600 done, total 5600/11522, average 0.099 sec/video, moving Prec@1 44.439 Prec@5 73.512
video 5800 done, total 5800/11522, average 0.099 sec/video, moving Prec@1 44.389 Prec@5 73.339
video 6000 done, total 6000/11522, average 0.098 sec/video, moving Prec@1 44.476 Prec@5 73.494
video 6200 done, total 6200/11522, average 0.098 sec/video, moving Prec@1 44.477 Prec@5 73.333
video 6400 done, total 6400/11522, average 0.098 sec/video, moving Prec@1 44.524 Prec@5 73.526
video 6600 done, total 6600/11522, average 0.097 sec/video, moving Prec@1 44.493 Prec@5 73.510
video 6800 done, total 6800/11522, average 0.097 sec/video, moving Prec@1 44.508 Prec@5 73.465
video 7000 done, total 7000/11522, average 0.096 sec/video, moving Prec@1 44.479 Prec@5 73.466
video 7200 done, total 7200/11522, average 0.096 sec/video, moving Prec@1 44.466 Prec@5 73.412
video 7400 done, total 7400/11522, average 0.096 sec/video, moving Prec@1 44.467 Prec@5 73.387
video 7600 done, total 7600/11522, average 0.096 sec/video, moving Prec@1 44.507 Prec@5 73.443
video 7800 done, total 7800/11522, average 0.096 sec/video, moving Prec@1 44.469 Prec@5 73.444
video 8000 done, total 8000/11522, average 0.095 sec/video, moving Prec@1 44.444 Prec@5 73.496
video 8200 done, total 8200/11522, average 0.095 sec/video, moving Prec@1 44.434 Prec@5 73.423
video 8400 done, total 8400/11522, average 0.094 sec/video, moving Prec@1 44.566 Prec@5 73.508
video 8600 done, total 8600/11522, average 0.094 sec/video, moving Prec@1 44.390 Prec@5 73.473
video 8800 done, total 8800/11522, average 0.094 sec/video, moving Prec@1 44.563 Prec@5 73.496
video 9000 done, total 9000/11522, average 0.093 sec/video, moving Prec@1 44.584 Prec@5 73.496
video 9200 done, total 9200/11522, average 0.092 sec/video, moving Prec@1 44.680 Prec@5 73.496
video 9400 done, total 9400/11522, average 0.092 sec/video, moving Prec@1 44.644 Prec@5 73.464
video 9600 done, total 9600/11522, average 0.092 sec/video, moving Prec@1 44.776 Prec@5 73.569
video 9800 done, total 9800/11522, average 0.092 sec/video, moving Prec@1 44.801 Prec@5 73.649
video 10000 done, total 10000/11522, average 0.091 sec/video, moving Prec@1 44.775 Prec@5 73.666
video 10200 done, total 10200/11522, average 0.091 sec/video, moving Prec@1 44.848 Prec@5 73.761
video 10400 done, total 10400/11522, average 0.091 sec/video, moving Prec@1 44.861 Prec@5 73.737
video 10600 done, total 10600/11522, average 0.091 sec/video, moving Prec@1 44.882 Prec@5 73.723
video 10800 done, total 10800/11522, average 0.091 sec/video, moving Prec@1 44.986 Prec@5 73.747
video 11000 done, total 11000/11522, average 0.090 sec/video, moving Prec@1 44.932 Prec@5 73.706
video 11200 done, total 11200/11522, average 0.090 sec/video, moving Prec@1 44.996 Prec@5 73.747
video 11400 done, total 11400/11522, average 0.090 sec/video, moving Prec@1 44.978 Prec@5 73.699
[0.7761194 0.275 0.19047619 0.62686567 0.23684211 0.40517241
0.62686567 0.44117647 0.45631068 0.6407767 0.37333333 0.38888889
0.24468085 0.32258065 0.7133758 0.36 0.19879518 0.225
0.39837398 0.28 0.32 0.48275862 0.40740741 0.43137255
0.32258065 0.33333333 0.39130435 0.28 0.60606061 0.53968254
0.63636364 0.57692308 0.72222222 0.10989011 0.23913043 0.16666667
0.7816092 0.78431373 0.05263158 0.48148148 0.62809917 0.54
0.66923077 0.63265306 0.59615385 0.52459016 0.41176471 0.34736842
0.15730337 0.66165414 0.56603774 0. 0.36 0.4
0.19047619 0.29166667 0.39393939 0.40625 0. 0.73469388
0.60606061 0.13888889 0.6 0.11818182 0.1 0.38888889
0.11111111 0.15151515 0.43661972 0.30769231 0.48387097 0.25
0.5 0.44 0.7 0.41935484 0.29411765 0.29411765
0.13636364 0.63414634 0.11428571 0.14285714 0.33333333 0.10526316
0.62857143 0.29310345 0.52702703 0.55555556 0.16666667 0.10714286
0.48648649 0.68292683 0.41860465 0.46296296 0.57009346 0.27419355
0.03703704 0.11904762 0.41520468 0.41176471 0.24489796 0.51851852
0.21875 0.46938776 0.42682927 0.6 0.5505618 0.75362319
0.47368421 0.2967033 0.54901961 0.11764706 0.60810811 0.38095238
0.66037736 0.36666667 0.21818182 0.62765957 0.2195122 0.58333333
0.79545455 0.54901961 0.35454545 0.425 0.36111111 0.44897959
0.49056604 0.53703704 0.15862069 0.53846154 0.22 0.18518519
0.13157895 0.23333333 0.64238411 0.29310345 0.07142857 0.45
0.08474576 0.51315789 0.57 0.34883721 0.25 0.38135593
0.56 0.33098592 0.6509434 0.17910448 0.484375 0.752
0.56363636 0.12903226 0.28333333 0.7037037 0.5 0.15686275
0.18032787 0.21428571 0.43564356 0.4375 0.24074074 0.25
0.26 0.31034483 0.60784314 0.66666667 0.83495146 0.75471698
0.68518519 0.45945946 0.44761905 0.61870504 0.59124088 0.71052632]
upper bound: 0.42524204990350917
-----Evaluation is finished------
Class Accuracy 40.94%
Overall Prec@1 44.97% Prec@5 73.69%
E:\Program Files\Anaconda3\envs\tsm\lib\site-packages\numpy\core\_asarray.py:136: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray return array(a, dtype, copy=False, order=order, subok=True)