YOLO

This repo is based on Jittor. It supports yolov3 and yolov5.

COCO128 datasets: https://cloud.tsinghua.edu.cn/f/7efbb154e31449d78e2b/

Test Results

Pytorch(coco128)

Model	Class	Images	Targets	P	R	[email protected]	[email protected]:.95
yolov3-spp	all	128	929	0.549	0.868	0.82	0.574
yolov3	all	128	929	0.527	0.831	0.791	0.56
yolov3-tiny	all	128	929	0.399	0.435	0.442	0.229
yolov5s	all	128	929	0.641	0.637	0.656	0.429
yolov5m	all	128	929	0.775	0.702	0.788	0.569
yolov5l	all	128	929	0.8	0.743	0.821	0.614
yolov5x	all	128	929	0.829	0.757	0.857	0.642

Jittor(coco128)

Model	Class	Images	Targets	P	R	[email protected]	[email protected]:.95
yolov3-spp	all	128	929	0.551	0.869	0.821	0.574
yolov3	all	128	929	0.53	0.829	0.79	0.561
yolov3-tiny	all	128	929	0.454	0.44	0.45	0.234
yolov5s	all	128	929	0.639	0.637	0.655	0.427
yolov5m	all	128	929	0.769	0.707	0.787	0.569
yolov5l	all	128	929	0.801	0.742	0.821	0.613
yolov5x	all	128	929	0.83	0.756	0.857	0.642

All Checkpoint links is here

Reference

1. https://github.com/Jittor/jittor
2. https://github.com/ultralytics/yolov3
3. https://github.com/ultralytics/yolov5