model cfg, weight parser from Tianxiaomo / pytorch-YOLOv4
evaluation에 사용되는 coco dataset은 test_parsing.py 내에 class: opt로 하드코딩 되어있음
default는 coco val2017 이미지 64개
class: opt:
anno_json = '/workspace/GitHub/YOLO/coco_forTest/annotations/instances_val2017_64.json'
pred_json = './YOLOv4_pred_parserTest.json'
img_path = '/workspace/GitHub/YOLO/coco_forTest/images/val2017_64/'결과는 Model_Parsing_Eval.ipynb 참조
example output (yolov4)
$ python test_parsing.py \
-cfgfile '/workspace/GitHub/YOLO/cfg/yolov4.cfg' \
-weights '/workspace/GitHub/YOLO/weights/yolov4.weights'
convalution havn't activate linear
convalution havn't activate linear
convalution havn't activate linear
layer filters size input output
0 conv 32 3 x 3 / 1 608 x 608 x 3 -> 608 x 608 x 32
1 conv 64 3 x 3 / 2 608 x 608 x 32 -> 304 x 304 x 64
2 conv 64 1 x 1 / 1 304 x 304 x 64 -> 304 x 304 x 64
3 route 1
4 conv 64 1 x 1 / 1 304 x 304 x 64 -> 304 x 304 x 64
5 conv 32 1 x 1 / 1 304 x 304 x 64 -> 304 x 304 x 32
6 conv 64 3 x 3 / 1 304 x 304 x 32 -> 304 x 304 x 64
7 shortcut 4
8 conv 64 1 x 1 / 1 304 x 304 x 64 -> 304 x 304 x 64
9 route 8 2
10 conv 64 1 x 1 / 1 304 x 304 x 128 -> 304 x 304 x 64
11 conv 128 3 x 3 / 2 304 x 304 x 64 -> 152 x 152 x 128
12 conv 64 1 x 1 / 1 152 x 152 x 128 -> 152 x 152 x 64
13 route 11
14 conv 64 1 x 1 / 1 152 x 152 x 128 -> 152 x 152 x 64
15 conv 64 1 x 1 / 1 152 x 152 x 64 -> 152 x 152 x 64
16 conv 64 3 x 3 / 1 152 x 152 x 64 -> 152 x 152 x 64
17 shortcut 14
18 conv 64 1 x 1 / 1 152 x 152 x 64 -> 152 x 152 x 64
19 conv 64 3 x 3 / 1 152 x 152 x 64 -> 152 x 152 x 64
20 shortcut 17
21 conv 64 1 x 1 / 1 152 x 152 x 64 -> 152 x 152 x 64
22 route 21 12
23 conv 128 1 x 1 / 1 152 x 152 x 128 -> 152 x 152 x 128
24 conv 256 3 x 3 / 2 152 x 152 x 128 -> 76 x 76 x 256
25 conv 128 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 128
26 route 24
27 conv 128 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 128
28 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
29 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
30 shortcut 27
31 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
32 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
33 shortcut 30
34 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
35 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
36 shortcut 33
37 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
38 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
39 shortcut 36
40 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
41 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
42 shortcut 39
43 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
44 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
45 shortcut 42
46 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
47 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
48 shortcut 45
49 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
50 conv 128 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 128
51 shortcut 48
52 conv 128 1 x 1 / 1 76 x 76 x 128 -> 76 x 76 x 128
53 route 52 25
54 conv 256 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 256
55 conv 512 3 x 3 / 2 76 x 76 x 256 -> 38 x 38 x 512
56 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
57 route 55
58 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
59 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
60 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
61 shortcut 58
62 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
63 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
64 shortcut 61
65 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
66 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
67 shortcut 64
68 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
69 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
70 shortcut 67
71 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
72 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
73 shortcut 70
74 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
75 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
76 shortcut 73
77 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
78 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
79 shortcut 76
80 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
81 conv 256 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 256
82 shortcut 79
83 conv 256 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 256
84 route 83 56
85 conv 512 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 512
86 conv 1024 3 x 3 / 2 38 x 38 x 512 -> 19 x 19 x1024
87 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
88 route 86
89 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
90 conv 512 1 x 1 / 1 19 x 19 x 512 -> 19 x 19 x 512
91 conv 512 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x 512
92 shortcut 89
93 conv 512 1 x 1 / 1 19 x 19 x 512 -> 19 x 19 x 512
94 conv 512 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x 512
95 shortcut 92
96 conv 512 1 x 1 / 1 19 x 19 x 512 -> 19 x 19 x 512
97 conv 512 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x 512
98 shortcut 95
99 conv 512 1 x 1 / 1 19 x 19 x 512 -> 19 x 19 x 512
100 conv 512 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x 512
101 shortcut 98
102 conv 512 1 x 1 / 1 19 x 19 x 512 -> 19 x 19 x 512
103 route 102 87
104 conv 1024 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x1024
105 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
106 conv 1024 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x1024
107 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
108 max 5 x 5 / 1 19 x 19 x 512 -> 19 x 19 x 512
109 route 107
110 max 9 x 9 / 1 19 x 19 x 512 -> 19 x 19 x 512
111 route 107
112 max 13 x 13 / 1 19 x 19 x 512 -> 19 x 19 x 512
113 route 112 110 108 107
114 conv 512 1 x 1 / 1 19 x 19 x2048 -> 19 x 19 x 512
115 conv 1024 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x1024
116 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
117 conv 256 1 x 1 / 1 19 x 19 x 512 -> 19 x 19 x 256
118 upsample * 2 19 x 19 x 256 -> 38 x 38 x 256
119 route 85
120 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
121 route 120 118
122 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
123 conv 512 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 512
124 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
125 conv 512 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 512
126 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
127 conv 128 1 x 1 / 1 38 x 38 x 256 -> 38 x 38 x 128
128 upsample * 2 38 x 38 x 128 -> 76 x 76 x 128
129 route 54
130 conv 128 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 128
131 route 130 128
132 conv 128 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 128
133 conv 256 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 256
134 conv 128 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 128
135 conv 256 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 256
136 conv 128 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 128
137 conv 256 3 x 3 / 1 76 x 76 x 128 -> 76 x 76 x 256
138 conv 255 1 x 1 / 1 76 x 76 x 256 -> 76 x 76 x 255
139 detection
140 route 136
141 conv 256 3 x 3 / 2 76 x 76 x 128 -> 38 x 38 x 256
142 route 141 126
143 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
144 conv 512 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 512
145 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
146 conv 512 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 512
147 conv 256 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 256
148 conv 512 3 x 3 / 1 38 x 38 x 256 -> 38 x 38 x 512
149 conv 255 1 x 1 / 1 38 x 38 x 512 -> 38 x 38 x 255
150 detection
151 route 147
152 conv 512 3 x 3 / 2 38 x 38 x 256 -> 19 x 19 x 512
153 route 152 116
154 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
155 conv 1024 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x1024
156 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
157 conv 1024 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x1024
158 conv 512 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 512
159 conv 1024 3 x 3 / 1 19 x 19 x 512 -> 19 x 19 x1024
160 conv 255 1 x 1 / 1 19 x 19 x1024 -> 19 x 19 x 255
161 detection
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
STRAT detection
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 16/16 [00:01<00:00, 9.29it/s]
Convert results to COCO format
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 64/64 [00:00<00:00, 704.63it/s]
Loading and preparing results...
DONE (t=0.05s)
creating index...
index created!
Running per image evaluation...
Evaluate annotation type *bbox*
DONE (t=0.87s).
Accumulating evaluation results...
DONE (t=0.24s).
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.512
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.725
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.585
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.300
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.632
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.754
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.378
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.562
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.589
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.375
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.676
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.811
example output (yolov4-tiny)
python test_parsing.py \
-cfgfile '/workspace/GitHub/YOLO/cfg/yolov4-tiny.cfg' \
-weights '/workspace/GitHub/YOLO/weights/yolov4-tiny.weights'
convalution havn't activate linear
convalution havn't activate linear
layer filters size input output
0 conv 32 3 x 3 / 2 416 x 416 x 3 -> 208 x 208 x 32
1 conv 64 3 x 3 / 2 208 x 208 x 32 -> 104 x 104 x 64
2 conv 64 3 x 3 / 1 104 x 104 x 64 -> 104 x 104 x 64
3 route 2
4 conv 32 3 x 3 / 1 104 x 104 x 64 -> 104 x 104 x 32
5 conv 32 3 x 3 / 1 104 x 104 x 32 -> 104 x 104 x 32
6 route 5 4
7 conv 64 1 x 1 / 1 104 x 104 x 64 -> 104 x 104 x 64
8 route 2 7
9 max 2 x 2 / 2 104 x 104 x 128 -> 52 x 52 x 64
10 conv 128 3 x 3 / 1 52 x 52 x 64 -> 52 x 52 x 128
11 route 10
12 conv 64 3 x 3 / 1 52 x 52 x 128 -> 52 x 52 x 64
13 conv 64 3 x 3 / 1 52 x 52 x 64 -> 52 x 52 x 64
14 route 13 12
15 conv 128 1 x 1 / 1 52 x 52 x 128 -> 52 x 52 x 128
16 route 10 15
17 max 2 x 2 / 2 52 x 52 x 256 -> 26 x 26 x 128
18 conv 256 3 x 3 / 1 26 x 26 x 128 -> 26 x 26 x 256
19 route 18
20 conv 128 3 x 3 / 1 26 x 26 x 256 -> 26 x 26 x 128
21 conv 128 3 x 3 / 1 26 x 26 x 128 -> 26 x 26 x 128
22 route 21 20
23 conv 256 1 x 1 / 1 26 x 26 x 256 -> 26 x 26 x 256
24 route 18 23
25 max 2 x 2 / 2 26 x 26 x 512 -> 13 x 13 x 256
26 conv 512 3 x 3 / 1 13 x 13 x 256 -> 13 x 13 x 512
27 conv 256 1 x 1 / 1 13 x 13 x 512 -> 13 x 13 x 256
28 conv 512 3 x 3 / 1 13 x 13 x 256 -> 13 x 13 x 512
29 conv 255 1 x 1 / 1 13 x 13 x 512 -> 13 x 13 x 255
30 detection
31 route 27
32 conv 128 1 x 1 / 1 13 x 13 x 256 -> 13 x 13 x 128
33 upsample * 2 13 x 13 x 128 -> 26 x 26 x 128
34 route 33 23
35 conv 256 3 x 3 / 1 26 x 26 x 384 -> 26 x 26 x 256
36 conv 255 1 x 1 / 1 26 x 26 x 256 -> 26 x 26 x 255
37 detection
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
STRAT detection
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 16/16 [00:00<00:00, 19.44it/s]
Convert results to COCO format
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 64/64 [00:00<00:00, 817.01it/s]
Loading and preparing results...
DONE (t=0.04s)
creating index...
index created!
Running per image evaluation...
Evaluate annotation type *bbox*
DONE (t=0.71s).
Accumulating evaluation results...
DONE (t=0.23s).
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.303
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.518
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.309
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.092
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.417
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.433
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.278
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.384
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.398
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.125
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.517
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.56
pytorch -> onnx 는 demo_pytorch2onnx.py 참조
onnx -> tensorrt 는 https://github.com/khlee369/Pytorch2TensorRT 참조