Skip to content

Commit 364fcfd

Browse files
glenn-jocherglenn-jocher
committed
PANet update
1 parent ef58dac commit 364fcfd

File tree

7 files changed

+126
-96
lines changed

7 files changed

+126
-96
lines changed

‎README.md

Lines changed: 8 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44

55
This repository represents Ultralytics open-source research into future object detection methods, and incorporates our lessons learned and best practices evolved over training thousands of models on custom client datasets with our previous YOLO repository https://github.com/ultralytics/yolov3. **All code and models are under active development, and are subject to modification or deletion without notice.** Use at your own risk.
66

7-
<img src="https://user-images.githubusercontent.com/26833433/84200349-729f2680-aa5b-11ea-8f9a-604c9e01a658.png" width="1000">** GPU Speed measures end-to-end time per image averaged over 5000 COCO val2017 images using a V100 GPU with batch size 32, and includes image preprocessing, PyTorch FP32 inference, postprocessing and NMS.
7+
<img src="https://user-images.githubusercontent.com/26833433/85336627-c6663280-b493-11ea-9b0a-289b0f182b84.png" width="1000">** GPU Speed measures end-to-end time per image averaged over 5000 COCO val2017 images using a V100 GPU with batch size 8, and includes image preprocessing, PyTorch FP16 inference, postprocessing and NMS.
88

99
- **June 19, 2020**: [FP16](https://pytorch.org/docs/stable/nn.html#torch.nn.Module.half) as new default for smaller checkpoints and faster inference. Comparison in [d4c6674](https://github.com/ultralytics/yolov5/commit/d4c6674c98e19df4c40e33a777610a18d1961145).
1010
- **June 9, 2020**: [CSP](https://github.com/WongKinYiu/CrossStagePartialNetworks) updates to all YOLOv5 models. New models are faster, smaller and more accurate. Credit to @WongKinYiu for his excellent work with CSP.
@@ -14,13 +14,14 @@ This repository represents Ultralytics open-source research into future object d
1414

1515
## Pretrained Checkpoints
1616

17-
| Model | AP<sup>val</sup> | AP<sup>test</sup> | AP<sub>50</sub> | Speed<sub>GPU</sub> | FPS<sub>GPU</sub> || params | FLOPs |
17+
| Model | AP<sup>val</sup> | AP<sup>test</sup> | AP<sub>50</sub> | Speed<sub>GPU</sub> | FPS<sub>GPU</sub> || params | FLOPS |
1818
|---------- |------ |------ |------ | -------- | ------| ------ |------ | :------: |
19-
| YOLOv5-s ([ckpt](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J)) | 35.5 | 35.5 | 55.0 | **2.1ms** | **476** || 7.1M | 12.6B
20-
| YOLOv5-m ([ckpt](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J)) | 42.7 | 42.7 | 62.4 | 3.2ms | 312 || 22.0M | 39.0B
21-
| YOLOv5-l ([ckpt](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J)) | 45.7 | 45.9 | 65.1 | 4.1ms | 243 || 50.3M | 89.0B
22-
| YOLOv5-x ([ckpt](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J)) | **47.2** | **47.3** | **66.6** | 6.5ms | 153 || 95.9M | 170.3B
23-
| YOLOv3-SPP ([ckpt](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J)) | 45.6 | 45.5 | 65.2 | 4.8ms | 208 || 63.0M | 118.0B
19+
| [YOLOv5s](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J) | 36.5 | 36.5 | 55.6 | **2.2ms** | **455** || 7.5M | 13.2B
20+
| [YOLOv5m](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J) | 43.4 | 43.4 | 62.4 | 3.0ms | 333 || 21.8M | 39.4B
21+
| [YOLOv5l](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J) | 46.6 | 46.7 | 65.4 | 3.9ms | 256 || 47.8M | 88.1B
22+
| [YOLOv5x](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J) | **48.2** | **48.3** | **66.9** | 6.1ms | 164 || 89.0M | 166.4B
23+
| [YOLOv3-SPP](https://drive.google.com/open?id=1Drs_Aiu7xx6S-ix95f9kNsA6ueKRpN2J) | 45.6 | 45.5 | 65.2 | 4.5ms | 222 || 63.0M | 118.0B
24+
2425

2526
** AP<sup>test</sup> denotes COCO [test-dev2017](http://cocodataset.org/#upload) server results, all other AP results in the table denote val2017 accuracy.
2627
** All AP numbers are for single-model single-scale without ensemble or test-time augmentation. Reproduce by `python test.py --img 736 --conf 0.001`

‎models/yolov3-spp.yaml

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -25,8 +25,7 @@ backbone:
2525
[-1, 4, Bottleneck, [1024]], # 10
2626
]
2727

28-
# yolov3-spp head
29-
# na = len(anchors[0])
28+
# YOLOv3-SPP head
3029
head:
3130
[[-1, 1, Bottleneck, [1024, False]], # 11
3231
[-1, 1, SPP, [512, [5, 9, 13]]],

‎models/yolov5l.yaml

Lines changed: 28 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -5,41 +5,48 @@ width_multiple: 1.0 # layer channel multiple
55

66
# anchors
77
anchors:
8-
- [10,13, 16,30, 33,23] # P3/8
9-
- [30,61, 62,45, 59,119] # P4/16
108
- [116,90, 156,198, 373,326] # P5/32
9+
- [30,61, 62,45, 59,119] # P4/16
10+
- [10,13, 16,30, 33,23] # P3/8
1111

12-
# yolov5 backbone
12+
# YOLOv5 backbone
1313
backbone:
1414
# [from, number, module, args]
15-
[[-1, 1, Focus, [64, 3]], # 1-P1/2
16-
[-1, 1, Conv, [128, 3, 2]], # 2-P2/4
17-
[-1, 3, Bottleneck, [128]],
18-
[-1, 1, Conv, [256, 3, 2]], # 4-P3/8
15+
[[-1, 1, Focus, [64, 3]], # 0-P1/2
16+
[-1, 1, Conv, [128, 3, 2]], # 1-P2/4
17+
[-1, 3, BottleneckCSP, [128]],
18+
[-1, 1, Conv, [256, 3, 2]], # 3-P3/8
1919
[-1, 9, BottleneckCSP, [256]],
20-
[-1, 1, Conv, [512, 3, 2]], # 6-P4/16
20+
[-1, 1, Conv, [512, 3, 2]], # 5-P4/16
2121
[-1, 9, BottleneckCSP, [512]],
22-
[-1, 1, Conv, [1024, 3, 2]], # 8-P5/32
22+
[-1, 1, Conv, [1024, 3, 2]], # 7-P5/32
2323
[-1, 1, SPP, [1024, [5, 9, 13]]],
24-
[-1, 6, BottleneckCSP, [1024]], # 10
2524
]
2625

27-
# yolov5 head
26+
# YOLOv5 head
2827
head:
29-
[[-1, 3, BottleneckCSP, [1024, False]], # 11
30-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 12 (P5/32-large)
28+
[[-1, 3, BottleneckCSP, [1024, False]], # 9
3129

32-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
33-
[[-1, 6], 1, Concat, [1]], # cat backbone P4
3430
[-1, 1, Conv, [512, 1, 1]],
35-
[-1, 3, BottleneckCSP, [512, False]],
36-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 17 (P4/16-medium)
31+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
32+
[[-1, 6], 1, Concat, [1]], # cat backbone P4
33+
[-1, 3, BottleneckCSP, [512, False]], # 13
3734

38-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
39-
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4035
[-1, 1, Conv, [256, 1, 1]],
36+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
37+
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4138
[-1, 3, BottleneckCSP, [256, False]],
42-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P3/8-small)
39+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 18 (P3/8-small)
40+
41+
[-2, 1, Conv, [256, 3, 2]],
42+
[[-1, 14], 1, Concat, [1]], # cat head P4
43+
[-1, 3, BottleneckCSP, [512, False]],
44+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P4/16-medium)
45+
46+
[-2, 1, Conv, [512, 3, 2]],
47+
[[-1, 10], 1, Concat, [1]], # cat head P5
48+
[-1, 3, BottleneckCSP, [1024, False]],
49+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 26 (P5/32-large)
4350

44-
[[], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
51+
[[], 1, Detect, [nc, anchors]], # Detect(P5, P4, P3)
4552
]

‎models/yolov5m.yaml

Lines changed: 28 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -5,41 +5,48 @@ width_multiple: 0.75 # layer channel multiple
55

66
# anchors
77
anchors:
8-
- [10,13, 16,30, 33,23] # P3/8
9-
- [30,61, 62,45, 59,119] # P4/16
108
- [116,90, 156,198, 373,326] # P5/32
9+
- [30,61, 62,45, 59,119] # P4/16
10+
- [10,13, 16,30, 33,23] # P3/8
1111

12-
# yolov5 backbone
12+
# YOLOv5 backbone
1313
backbone:
1414
# [from, number, module, args]
15-
[[-1, 1, Focus, [64, 3]], # 1-P1/2
16-
[-1, 1, Conv, [128, 3, 2]], # 2-P2/4
17-
[-1, 3, Bottleneck, [128]],
18-
[-1, 1, Conv, [256, 3, 2]], # 4-P3/8
15+
[[-1, 1, Focus, [64, 3]], # 0-P1/2
16+
[-1, 1, Conv, [128, 3, 2]], # 1-P2/4
17+
[-1, 3, BottleneckCSP, [128]],
18+
[-1, 1, Conv, [256, 3, 2]], # 3-P3/8
1919
[-1, 9, BottleneckCSP, [256]],
20-
[-1, 1, Conv, [512, 3, 2]], # 6-P4/16
20+
[-1, 1, Conv, [512, 3, 2]], # 5-P4/16
2121
[-1, 9, BottleneckCSP, [512]],
22-
[-1, 1, Conv, [1024, 3, 2]], # 8-P5/32
22+
[-1, 1, Conv, [1024, 3, 2]], # 7-P5/32
2323
[-1, 1, SPP, [1024, [5, 9, 13]]],
24-
[-1, 6, BottleneckCSP, [1024]], # 10
2524
]
2625

27-
# yolov5 head
26+
# YOLOv5 head
2827
head:
29-
[[-1, 3, BottleneckCSP, [1024, False]], # 11
30-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 12 (P5/32-large)
28+
[[-1, 3, BottleneckCSP, [1024, False]], # 9
3129

32-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
33-
[[-1, 6], 1, Concat, [1]], # cat backbone P4
3430
[-1, 1, Conv, [512, 1, 1]],
35-
[-1, 3, BottleneckCSP, [512, False]],
36-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 17 (P4/16-medium)
31+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
32+
[[-1, 6], 1, Concat, [1]], # cat backbone P4
33+
[-1, 3, BottleneckCSP, [512, False]], # 13
3734

38-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
39-
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4035
[-1, 1, Conv, [256, 1, 1]],
36+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
37+
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4138
[-1, 3, BottleneckCSP, [256, False]],
42-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P3/8-small)
39+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 18 (P3/8-small)
40+
41+
[-2, 1, Conv, [256, 3, 2]],
42+
[[-1, 14], 1, Concat, [1]], # cat head P4
43+
[-1, 3, BottleneckCSP, [512, False]],
44+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P4/16-medium)
45+
46+
[-2, 1, Conv, [512, 3, 2]],
47+
[[-1, 10], 1, Concat, [1]], # cat head P5
48+
[-1, 3, BottleneckCSP, [1024, False]],
49+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 26 (P5/32-large)
4350

44-
[[], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
51+
[[], 1, Detect, [nc, anchors]], # Detect(P5, P4, P3)
4552
]

‎models/yolov5s.yaml

Lines changed: 28 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -5,41 +5,48 @@ width_multiple: 0.50 # layer channel multiple
55

66
# anchors
77
anchors:
8-
- [10,13, 16,30, 33,23] # P3/8
9-
- [30,61, 62,45, 59,119] # P4/16
108
- [116,90, 156,198, 373,326] # P5/32
9+
- [30,61, 62,45, 59,119] # P4/16
10+
- [10,13, 16,30, 33,23] # P3/8
1111

12-
# yolov5 backbone
12+
# YOLOv5 backbone
1313
backbone:
1414
# [from, number, module, args]
15-
[[-1, 1, Focus, [64, 3]], # 1-P1/2
16-
[-1, 1, Conv, [128, 3, 2]], # 2-P2/4
17-
[-1, 3, Bottleneck, [128]],
18-
[-1, 1, Conv, [256, 3, 2]], # 4-P3/8
15+
[[-1, 1, Focus, [64, 3]], # 0-P1/2
16+
[-1, 1, Conv, [128, 3, 2]], # 1-P2/4
17+
[-1, 3, BottleneckCSP, [128]],
18+
[-1, 1, Conv, [256, 3, 2]], # 3-P3/8
1919
[-1, 9, BottleneckCSP, [256]],
20-
[-1, 1, Conv, [512, 3, 2]], # 6-P4/16
20+
[-1, 1, Conv, [512, 3, 2]], # 5-P4/16
2121
[-1, 9, BottleneckCSP, [512]],
22-
[-1, 1, Conv, [1024, 3, 2]], # 8-P5/32
22+
[-1, 1, Conv, [1024, 3, 2]], # 7-P5/32
2323
[-1, 1, SPP, [1024, [5, 9, 13]]],
24-
[-1, 6, BottleneckCSP, [1024]], # 10
2524
]
2625

27-
# yolov5 head
26+
# YOLOv5 head
2827
head:
29-
[[-1, 3, BottleneckCSP, [1024, False]], # 11
30-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 12 (P5/32-large)
28+
[[-1, 3, BottleneckCSP, [1024, False]], # 9
3129

32-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
33-
[[-1, 6], 1, Concat, [1]], # cat backbone P4
3430
[-1, 1, Conv, [512, 1, 1]],
35-
[-1, 3, BottleneckCSP, [512, False]],
36-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 17 (P4/16-medium)
31+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
32+
[[-1, 6], 1, Concat, [1]], # cat backbone P4
33+
[-1, 3, BottleneckCSP, [512, False]], # 13
3734

38-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
39-
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4035
[-1, 1, Conv, [256, 1, 1]],
36+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
37+
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4138
[-1, 3, BottleneckCSP, [256, False]],
42-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P3/8-small)
39+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 18 (P3/8-small)
40+
41+
[-2, 1, Conv, [256, 3, 2]],
42+
[[-1, 14], 1, Concat, [1]], # cat head P4
43+
[-1, 3, BottleneckCSP, [512, False]],
44+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P4/16-medium)
45+
46+
[-2, 1, Conv, [512, 3, 2]],
47+
[[-1, 10], 1, Concat, [1]], # cat head P5
48+
[-1, 3, BottleneckCSP, [1024, False]],
49+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 26 (P5/32-large)
4350

44-
[[], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
51+
[[], 1, Detect, [nc, anchors]], # Detect(P5, P4, P3)
4552
]

‎models/yolov5x.yaml

Lines changed: 28 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -5,41 +5,48 @@ width_multiple: 1.25 # layer channel multiple
55

66
# anchors
77
anchors:
8-
- [10,13, 16,30, 33,23] # P3/8
9-
- [30,61, 62,45, 59,119] # P4/16
108
- [116,90, 156,198, 373,326] # P5/32
9+
- [30,61, 62,45, 59,119] # P4/16
10+
- [10,13, 16,30, 33,23] # P3/8
1111

12-
# yolov5 backbone
12+
# YOLOv5 backbone
1313
backbone:
1414
# [from, number, module, args]
15-
[[-1, 1, Focus, [64, 3]], # 1-P1/2
16-
[-1, 1, Conv, [128, 3, 2]], # 2-P2/4
17-
[-1, 3, Bottleneck, [128]],
18-
[-1, 1, Conv, [256, 3, 2]], # 4-P3/8
15+
[[-1, 1, Focus, [64, 3]], # 0-P1/2
16+
[-1, 1, Conv, [128, 3, 2]], # 1-P2/4
17+
[-1, 3, BottleneckCSP, [128]],
18+
[-1, 1, Conv, [256, 3, 2]], # 3-P3/8
1919
[-1, 9, BottleneckCSP, [256]],
20-
[-1, 1, Conv, [512, 3, 2]], # 6-P4/16
20+
[-1, 1, Conv, [512, 3, 2]], # 5-P4/16
2121
[-1, 9, BottleneckCSP, [512]],
22-
[-1, 1, Conv, [1024, 3, 2]], # 8-P5/32
22+
[-1, 1, Conv, [1024, 3, 2]], # 7-P5/32
2323
[-1, 1, SPP, [1024, [5, 9, 13]]],
24-
[-1, 6, BottleneckCSP, [1024]], # 10
2524
]
2625

27-
# yolov5 head
26+
# YOLOv5 head
2827
head:
29-
[[-1, 3, BottleneckCSP, [1024, False]], # 11
30-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 12 (P5/32-large)
28+
[[-1, 3, BottleneckCSP, [1024, False]], # 9
3129

32-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
33-
[[-1, 6], 1, Concat, [1]], # cat backbone P4
3430
[-1, 1, Conv, [512, 1, 1]],
35-
[-1, 3, BottleneckCSP, [512, False]],
36-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 17 (P4/16-medium)
31+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
32+
[[-1, 6], 1, Concat, [1]], # cat backbone P4
33+
[-1, 3, BottleneckCSP, [512, False]], # 13
3734

38-
[-2, 1, nn.Upsample, [None, 2, 'nearest']],
39-
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4035
[-1, 1, Conv, [256, 1, 1]],
36+
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
37+
[[-1, 4], 1, Concat, [1]], # cat backbone P3
4138
[-1, 3, BottleneckCSP, [256, False]],
42-
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P3/8-small)
39+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 18 (P3/8-small)
40+
41+
[-2, 1, Conv, [256, 3, 2]],
42+
[[-1, 14], 1, Concat, [1]], # cat head P4
43+
[-1, 3, BottleneckCSP, [512, False]],
44+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 22 (P4/16-medium)
45+
46+
[-2, 1, Conv, [512, 3, 2]],
47+
[[-1, 10], 1, Concat, [1]], # cat head P5
48+
[-1, 3, BottleneckCSP, [1024, False]],
49+
[-1, 1, nn.Conv2d, [na * (nc + 5), 1, 1]], # 26 (P5/32-large)
4350

44-
[[], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
51+
[[], 1, Detect, [nc, anchors]], # Detect(P5, P4, P3)
4552
]

‎utils/utils.py

Lines changed: 5 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1094,12 +1094,14 @@ def plot_study_txt(f='study.txt', x=None): # from utils.utils import *; plot_st
10941094

10951095
ax2.plot(1E3 / np.array([209, 140, 97, 58, 35, 18]), [33.5, 39.1, 42.5, 45.9, 49., 50.5],
10961096
'k.-', linewidth=2, markersize=8, alpha=.25, label='EfficientDet')
1097+
1098+
ax2.grid()
10971099
ax2.set_xlim(0, 30)
1098-
ax2.set_ylim(25, 50)
1099-
ax2.set_xlabel('GPU Latency (ms)')
1100+
ax2.set_ylim(28, 50)
1101+
ax2.set_yticks(np.arange(30, 55, 5))
1102+
ax2.set_xlabel('GPU Speed (ms/img)')
11001103
ax2.set_ylabel('COCO AP val')
11011104
ax2.legend(loc='lower right')
1102-
ax2.grid()
11031105
plt.savefig('study_mAP_latency.png', dpi=300)
11041106
plt.savefig(f.replace('.txt', '.png'), dpi=200)
11051107

0 commit comments

Comments
 (0)