-
Notifications
You must be signed in to change notification settings - Fork 34
/
readme_yolox.txt
251 lines (162 loc) · 12.5 KB
/
readme_yolox.txt
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
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
# 如果命令不能成功执行，说明咩酱实现中，，，（and，把PPYOLOv2完全实现再补全文档)
nvidia-smi
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
wget https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_s.pth
wget https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_m.pth
wget https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_l.pth
wget https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_x.pth
----------------------- 转 ppdetection 的权重 -----------------------
python tools/convert_weights.py -f exps/yolox/yolox_s_voc2012.py -c 16.pdparams -oc pp_16.pth -nc 20
python tools/eval.py -f exps/yolox/yolox_s_voc2012.py -d 1 -b 8 -w 4 -c pp_16.pth --conf 0.01 --tsize 640
----------------------- 预处理 -----------------------
原版中YOLOX预处理：
1.读图片
img = cv2.imread(img)
2.填充成正方形，填充值是114,
if len(img.shape) == 3:
padded_img = np.ones((input_size[0], input_size[1], 3), dtype=np.uint8) * 114
else:
padded_img = np.ones(input_size, dtype=np.uint8) * 114
r = min(input_size[0] / img.shape[0], input_size[1] / img.shape[1])
3.保持宽高比把图片最长的边放缩到640，
resized_img = cv2.resize(
img,
(int(img.shape[1] * r), int(img.shape[0] * r)),
interpolation=cv2.INTER_LINEAR,
).astype(np.uint8)
padded_img[: int(img.shape[0] * r), : int(img.shape[1] * r)] = resized_img
4.transpose操作，HWC变成CHW
padded_img = padded_img.transpose(swap)
5.元素类型由int8转float32
padded_img = np.ascontiguousarray(padded_img, dtype=np.float32)
6.如果是旧版本，还需要这样（BGR转RGB，归一化）：
if self.legacy:
img = img[::-1, :, :].copy()
img /= 255.0
img -= np.array([0.485, 0.456, 0.406]).reshape(3, 1, 1)
img /= np.array([0.229, 0.224, 0.225]).reshape(3, 1, 1)
7.转Tensor:
img = torch.from_numpy(img).unsqueeze(0)
----------------------- 预测 -----------------------
python tools/demo.py image -f exps/yolox/yolox_s.py -c yolox_s.pth --path assets/dog.jpg --conf 0.25 --nms 0.45 --tsize 640 --save_result --device gpu
python tools/demo.py image -f exps/yolox/yolox_s.py -c yolox_s.pth --path assets/000000000019.jpg --conf 0.15 --tsize 640 --save_result --device gpu
python tools/demo.py image -f exps/yolox/yolox_m.py -c YOLOX_outputs/yolox_m/1.pth --path D://PycharmProjects/Paddle-PPYOLO-master/images/test --conf 0.25 --nms 0.45 --tsize 640 --save_result --device gpu
----------------------- 评估 -----------------------
python tools/eval.py -f exps/yolox/yolox_s.py -d 1 -b 8 -w 4 -c yolox_s.pth --conf 0.001 --tsize 640
Average forward time: 10.76 ms, Average NMS time: 2.75 ms, Average inference time: 13.51 ms
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.405
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.593
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.438
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.233
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.448
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.541
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.326
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.531
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.574
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.367
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.635
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.724
python tools/eval.py -f exps/yolox/yolox_m.py -d 1 -b 8 -w 4 -c yolox_m.pth --conf 0.001 --tsize 640
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.469
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.656
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.512
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.289
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.521
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.623
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.362
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.587
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.628
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.439
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.688
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.780
python tools/eval.py -f exps/yolox/yolox_l.py -d 1 -b 8 -w 4 -c yolox_l.pth --conf 0.001 --tsize 640
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.497
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.680
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.540
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.323
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.549
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.651
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 ] = 0.610
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.649
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.468
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.705
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.797
----------------------- 复现COCO上的精度 -----------------------
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
nohup python tools/train.py -f exps/yolox/yolox_s.py -d 8 -b 64 -eb 64 -w 4 -ew 4 --fp16 > yolox_s_coco.log 2>&1 &
nohup python tools/train.py -f exps/yolox/yolox_s.py -d 4 -b 64 -eb 64 -w 4 -ew 4 --fp16 > yolox_s_coco.log 2>&1 &
日志见 train_coco/yolox_s_8gpu.txt
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.399
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.588
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.431
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.220
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.443
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.529
python tools/eval.py -f exps/yolox/yolox_s.py -d 1 -b 16 -w 4 -c YOLOX_outputs/yolox_s/300.pth --conf 0.01 --tsize 640
(单卡调试用)
python tools/train.py -f exps/yolox/yolox_s.py -d 1 -b 8 -eb 4 -w 1 -ew 0 --fp16 -c yolox_s.pth
----------------------- 迁移学习，带上-c（--ckpt）参数读取预训练模型。 -----------------------
迁移学习（不冻结骨干网络）:（可以加--fp16， -eb表示验证时的批大小）
python tools/train.py -f exps/yolox/yolox_s_voc2012.py -d 1 -b 24 -eb 16 -w 4 -ew 4 -lrs 0.1 --fp16 -c yolox_s.pth
python tools/train.py -f exps/yolox/yolox_s_voc2012.py -d 1 -b 8 -eb 8 -w 1 -ew 0 -lrs 0.1 --fp16 -c yolox_s.pth
1机2卡训练：(发现一个隐藏知识点：获得损失（训练）、推理 都要放在模型的forward()中进行，否则DDP会计算错误结果。)
export CUDA_VISIBLE_DEVICES=0,1
nohup python tools/train.py -f exps/yolox/yolox_s_voc2012.py -d 2 -b 24 -eb 16 -w 4 -ew 4 -lrs 0.1 --fp16 -c yolox_s.pth > yolox_s.log 2>&1 &
python tools/eval.py -f exps/yolox/yolox_s_voc2012.py -d 1 -b 8 -w 4 -c 16.pth --conf 0.01 --tsize 640
实测 yolox_s 的AP最高可以到达（head.use_batch_assign = False,
日志见 train_ppyolo_in_voc2012/yolox_s_voc2012.txt 这是在完全空闲的服务器，2卡4090测速）
- - - - - - - - - - - -
export CUDA_VISIBLE_DEVICES=0,1
nohup python tools/train.py -f exps/yolox/yolox_l_voc2012.py -d 2 -b 24 -eb 16 -w 4 -ew 4 -lrs 0.1 --fp16 -c yolox_l.pth > yolox_l.log 2>&1 &
日志见 train_ppyolo_in_voc2012/yolox_l_voc2012.txt 这是在完全空闲的服务器，2卡4090测速）
- - - - - - - - - - - -
(迁移学习，发现使用更小的学习率才更好，所以指定 -lrs 参数。这个参数会乘到 basic_lr_per_img)
export CUDA_VISIBLE_DEVICES=0,1
nohup python tools/train.py -f exps/yolox/yolox_s_simple_voc2012.py -d 2 -b 24 -eb 16 -w 4 -ew 4 -lrs 0.1 --fp16 -c yolox_s.pth > yolox_s_simple.log 2>&1 &
tensorboard --logdir=./YOLOX_outputs/
实测 yolox_s_simple 的AP最后可以到达（head.use_batch_assign = False,
日志见 train_ppyolo_in_voc2012/yolox_s_simple_voc2012.txt 这是在完全空闲的服务器，2卡4090测速）
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.495
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.724
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.551
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.160
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.354
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.579
(测速。极速体验。)
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
nohup python tools/train.py -f exps/yolox/yolox_s_simple_voc2012.py -d 8 -b 64 -eb 64 -w 4 -ew 4 -lrs 0.1 --fp16 -c yolox_s.pth > yolox_s_simple_8gpu.log 2>&1 &
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5
nohup python tools/train.py -f exps/yolox/yolox_s_simple_voc2012.py -d 6 -b 48 -eb 48 -w 4 -ew 4 -lrs 0.1 --fp16 -c yolox_s.pth > yolox_s_simple.log 2>&1 &
export CUDA_VISIBLE_DEVICES=0,1,2,3
nohup python tools/train.py -f exps/yolox/yolox_s_simple_voc2012.py -d 4 -b 32 -eb 32 -w 4 -ew 4 -lrs 0.1 --fp16 -c yolox_s.pth > yolox_s_simple.log 2>&1 &
(单卡调试用)
python tools/train.py -f exps/yolox/yolox_s_simple_voc2012.py -d 1 -b 8 -eb 4 -w 1 -ew 0 --fp16 -c yolox_s.pth
python tools/eval.py -f exps/yolox/yolox_s_simple_voc2012.py -d 1 -b 8 -w 4 -c 16.pth --conf 0.01 --tsize 640
python tools/demo.py image -f exps/yolox/yolox_s_simple_voc2012.py -c 16.pth --path assets/2008_000073.jpg --conf 0.15 --tsize 640 --save_result --device gpu
导出给私有仓库：
python tools/convert_weights.py -f exps/yolox/yolox_s_simple_voc2012.py -c YOLOX_outputs/yolox_s_simple_voc2012/16.pth -oc new_16.pth -nc 20 -pp0
----------------------- 恢复训练（加上参数--resume） -----------------------
export CUDA_VISIBLE_DEVICES=0,1
nohup python tools/train.py -f exps/yolox/yolox_s_simple_voc2012.py -d 2 -b 24 -eb 16 -w 4 -ew 4 --fp16 -c 10.pth --resume > yolox_s_simple.log 2>&1 &
----------------------- 导出为ONNX -----------------------
见demo/ONNXRuntime/README.md
会设置model.head.decode_in_inference = False，此时只对置信位和各类别概率进行sigmoid()激活。xywh没有进行解码，更没有进行nms。
python tools/export_onnx.py --output-name yolox_s.onnx -f exps/yolox/yolox_s.py -c yolox_s.pth
ONNX预测，命令改动为（用numpy对xywh进行解码，进行nms。）
python tools/onnx_inference.py -m yolox_s.onnx -i assets/dog.jpg -o ONNX_YOLOX_outputs -s 0.3 --input_shape 640,640 -cn ./class_names/coco_classes.txt
用onnx模型进行验证，
python tools/onnx_eval.py -an YOLOX -m yolox_s.onnx -i ../COCO/val2017 -a ../COCO/annotations/instances_val2017.json -s 0.001 --input_shape 640,640 --eval_type eval
(不是因为少预测了没有gt的图片)
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.394
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.578
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.426
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.217
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.437
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.525
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.311
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.497
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.535
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.327
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.588
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.670