search

WongKinYiu / ScaledYOLOv4

Scaled-YOLOv4: Scaling Cross Stage Partial Network
GNU General Public License v3.0
2.02k stars 572 forks source link

Can you help me to mAP my custom dataset? #116

Open HansolEom opened 3 years ago

HansolEom commented 3 years ago

gt [gt]

pred [pred]

The shoes do not learn at all. I think small objects don't work well.

My training run: python -m torch.distributed.launch --nproc_per_node 4 train.py --batch-size 160 --data deep_moda.yaml --cfg yolov4-csp.cfg --weights '' --name yolov4-csp --sync-bn

Can you help me?

WongKinYiu commented 3 years ago

you could try to set image_weights=True in https://github.com/WongKinYiu/ScaledYOLOv4/blob/yolov4-csp/utils/datasets.py#L294.

HansolEom commented 3 years ago

you could try to set image_weights=True in https://github.com/WongKinYiu/ScaledYOLOv4/blob/yolov4-csp/utils/datasets.py#L294.

Thank you. I'll try. Right now, I am learning the same batch size you used. image The train obj_loss and val obj_loss are going in different directions. I think this problem can also be solved through image_weight.