ScaledYOLOv4 Download - ScaledYOLOv4 Source Code Download

Yolov4-large

هذا هو تطبيق "Scaled-Yolov4: Scaling Cross Stage Partial Network" باستخدام Pytorch Framwork.

YOLOV4-CSP
Yolov4-tiny
Yolov4-large

نموذج	حجم الاختبار	^اختبار AP	^اختبار AP ₅₀	AP ₇₅ ^اختبار	^اختبار AP _S.	^اختبار AP _M.	^اختبار AP _L.	PATCH1 إنتاجية
YOLOV4-P5	896	51.4 ٪	69.9 ٪	56.3 ٪	33.1 ٪	55.4 ٪	62.4 ٪	41 إطارًا في الثانية
YOLOV4-P5	تا	52.5 ٪	70.3 ٪	58.0 ٪	36.0 ٪	52.4 ٪	62.3 ٪	-

YOLOV4-P6	1280	54.3 ٪	72.3 ٪	59.5 ٪	36.6 ٪	58.2 ٪	65.5 ٪	30 إطارًا في الثانية
YOLOV4-P6	تا	54.9 ٪	72.6 ٪	60.2 ٪	37.4 ٪	58.8 ٪	66.7 ٪	-

Yolov4-P7	1536	55.4 ٪	73.3 ٪	60.7 ٪	38.1 ٪	59.5 ٪	67.4 ٪	15 إطارًا في الثانية
Yolov4-P7	تا	55.8 ٪	73.2 ٪	61.2 ٪	38.8 ٪	60.1 ٪	68.2 ٪	-

نموذج	حجم الاختبار	AP ^Val	AP ₅₀ ^فال	AP ₇₅ ^فال	AP _S ^Val	AP _M ^Val	AP _L ^Val	الأوزان
YOLOV4-P5	896	51.2 ٪	69.8 ٪	56.2 ٪	35.0 ٪	56.2 ٪	64.0 ٪	`yolov4-p5.pt`
YOLOV4-P5	تا	52.5 ٪	70.2 ٪	57.8 ٪	38.5 ٪	57.2 ٪	64.0 ٪	-
YOLOV4-P5 (+BOF)	896	51.7 ٪	70.3 ٪	56.7 ٪	35.9 ٪	56.7 ٪	64.3 ٪	`yolov4-p5_.pt`
YOLOV4-P5 (+BOF)	تا	52.8 ٪	70.6 ٪	58.3 ٪	38.8 ٪	57.4 ٪	64.4 ٪	-

YOLOV4-P6	1280	53.9 ٪	72.0 ٪	59.0 ٪	39.3 ٪	58.3 ٪	66.6 ٪	`yolov4-p6.pt`
YOLOV4-P6	تا	54.4 ٪	72.3 ٪	59.6 ٪	39.8 ٪	58.9 ٪	67.6 ٪	-
YOLOV4-P6 (+BOF)	1280	54.4 ٪	72.7 ٪	59.5 ٪	39.5 ٪	58.9 ٪	67.3 ٪	`yolov4-p6_.pt`
YOLOV4-P6 (+BOF)	تا	54.8 ٪	72.6 ٪	60.0 ٪	40.6 ٪	59.1 ٪	68.2 ٪	-
yolov4-p6 (+bof*)	1280	54.7 ٪	72.9 ٪	60.0 ٪	39.4 ٪	59.2 ٪	68.3 ٪
yolov4-p6 (+bof*)	تا	55.3 ٪	73.2 ٪	60.8 ٪	40.5 ٪	59.9 ٪	69.4 ٪	-

Yolov4-P7	1536	55.0 ٪	72.9 ٪	60.2 ٪	39.8 ٪	59.9 ٪	68.4 ٪	`yolov4-p7.pt`
Yolov4-P7	تا	55.5 ٪	72.9 ٪	60.8 ٪	41.1 ٪	60.3 ٪	68.9 ٪	-

نموذج	حجم الاختبار	AP ^Val	AP ₅₀ ^فال	AP ₇₅ ^فال	AP _S ^Val	AP _M ^Val	AP _L ^Val
YOLOV4-P6-الاهتمام	1280	54.3 ٪	72.3 ٪	59.6 ٪	38.7 ٪	58.9 ٪	66.6 ٪

تثبيت

 # create the docker container, you can change the share memory size if you have more.
nvidia-docker run --name yolov4_csp -it -v your_coco_path/:/coco/ -v your_code_path/:/yolo --shm-size=64g nvcr.io/nvidia/pytorch:20.06-py3

# install mish-cuda, if you use different pytorch version, you could try https://github.com/thomasbrandon/mish-cuda
cd /
git clone https://github.com/JunnYu/mish-cuda
cd mish-cuda
python setup.py build install

# go to code folder
cd /yolo

الاختبار

 # download {yolov4-p5.pt, yolov4-p6.pt, yolov4-p7.pt} and put them in /yolo/weights/ folder.
python test.py --img 896 --conf 0.001 --batch 8 --device 0 --data coco.yaml --weights weights/yolov4-p5.pt
python test.py --img 1280 --conf 0.001 --batch 8 --device 0 --data coco.yaml --weights weights/yolov4-p6.pt
python test.py --img 1536 --conf 0.001 --batch 8 --device 0 --data coco.yaml --weights weights/yolov4-p7.pt

سوف تحصل على النتائج التالية:

 # yolov4-p5
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.51244
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.69771
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.56180
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.35021
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.56247
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.63983
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.38530
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.64048
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.69801
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.55487
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.74368
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.82826

 # yolov4-p6
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.53857
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.72015
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.59025
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.39285
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.58283
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.66580
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.39552
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.66504
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.72141
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.59193
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.75844
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.83981

 # yolov4-p7
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.55046
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.72925
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.60224
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.39836
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.59854
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.68405
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.40256
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.66929
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.72943
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.59943
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.76873
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.84460

تمرين

نستخدم وحدات معالجة الرسومات المتعددة للتدريب. {YOLOV4-P5 ، YOLOV4-P6 ، YOLOV4-P7} استخدم دقة الإدخال {896 ، 1280 ، 1536} للتدريب على التوالي.

 # yolov4-p5
python -m torch.distributed.launch --nproc_per_node 4 train.py --batch-size 64 --img 896 896 --data coco.yaml --cfg yolov4-p5.yaml --weights '' --sync-bn --device 0,1,2,3 --name yolov4-p5
python -m torch.distributed.launch --nproc_per_node 4 train.py --batch-size 64 --img 896 896 --data coco.yaml --cfg yolov4-p5.yaml --weights 'runs/exp0_yolov4-p5/weights/last_298.pt' --sync-bn --device 0,1,2,3 --name yolov4-p5-tune --hyp 'data/hyp.finetune.yaml' --epochs 450 --resume

إذا كانت عملية التدريب الخاصة بك ، فهذا بسبب بق الحشرات. فقط Ctrl+C للتوقف عن التدريب واستئناف التدريب بواسطة:

 # yolov4-p5
python -m torch.distributed.launch --nproc_per_node 4 train.py --batch-size 64 --img 896 896 --data coco.yaml --cfg yolov4-p5.yaml --weights 'runs/exp0_yolov4-p5/weights/last.pt' --sync-bn --device 0,1,2,3 --name yolov4-p5 --resume

اقتباس

 @InProceedings{Wang_2021_CVPR,
    author    = {Wang, Chien-Yao and Bochkovskiy, Alexey and Liao, Hong-Yuan Mark},
    title     = {{Scaled-YOLOv4}: Scaling Cross Stage Partial Network},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2021},
    pages     = {13029-13038}
}