yolov3_pytorch

基于pytorch的yolov3

Training

** train your own datase

** --data-cfg='cfg/my.coco

** change data/coco.names

** --coco=False

** Two style:** 1:coco style(class x y w h) xywh is Normalized

         ** 2:my dataset style(class x1 y1 x2 y2) xyxy is pixel

** if your dataset like coco style,please set --coco=True

** if set your own dataset,please set --coco=False

Start Training: Run train.py to begin training after downloading COCO data with data/get_coco_dataset.sh.

Inference

Run detect.py to apply trained weights to an image, such as zidane.jpg from the data/samples folder:

YOLOv3: detect.py --cfg cfg/yolov3.cfg --weights weights/yolov3.pt

YOLOv3-tiny: detect.py --cfg cfg/yolov3-tiny.cfg --weights weights/yolov3-tiny.pt

Webcam

Run detect.py with webcam=True to show a live webcam feed.

Pretrained Weights

• Darknet *.weights format: https://pjreddie.com/media/files/yolov3.weights
• PyTorch *.pt format: https://drive.google.com/drive/folders/1uxgUBemJVw9wZsdpboYbzUN4bcRhsuAI

参考网址

• https://github.com/ultralytics/yolov3
• https://pjreddie.com/darknet/yolo/
• https://blog.csdn.net/leviopku/article/details/82660381