search

longcw / yolo2-pytorch

YOLOv2 in PyTorch
1.54k stars 421 forks source link

The issue about bbox loss and mAP when training my own data. #68

Open applerjunjun opened 6 years ago

applerjunjun commented 6 years ago

I only used one-class data to train the model and the change of loss is shown below.

1

Why is the bbox loss unable to converg? The class loss and IOU loss seems good to me.

When there is one class, what does mAP really mean? (The result shows 0.55) Also, is it better to use IOU instead of mAP to test the model?

chenruishuang commented 5 years ago

Can you tell me how do you modified the code to train your own data set? i change the classes, but it occurred index error..

chenruishuang commented 5 years ago

thanks a lot!!