Gold-YOLO : YOLO with Gather-and-Distribute Mechanism

Benchmark

benchmark and pretrained weights

install requirements

pip install -r requirements.txt

eval

1. change the data/coco.yaml to your data path

2. download the MindSpore Checkpoint from Benchmark to the pretrain fold

3. eval Gold-YOLO

   # eval Gold-YOLO-n
   python tools/eval.py --reproduce_640_eval --data data/coco.yaml --conf_file configs/gold_yolo-n.py --weights pretrain/Gold_n-dist-state_dict_ms.ckpt
   
   # eval Gold-YOLO-s
   python tools/eval.py --reproduce_640_eval --data data/coco.yaml --conf_file configs/gold_yolo-s.py --weights pretrain/Gold_s-pre-dist-state_dict_ms.ckpt
   
   # eval Gold-YOLO-m
   python tools/eval.py --reproduce_640_eval --data data/coco.yaml --conf_file configs/gold_yolo-m.py --weights pretrain/Gold_m-pre-dist-state_dict_ms.ckpt
   
   # eval Gold-YOLO-l
   python tools/eval.py --reproduce_640_eval --data data/coco.yaml --conf_file configs/gold_yolo-l.py --weights pretrain/Gold_l-pre-dist-state_dict_ms.ckpt