The loss function for the network

[zhangkuncsdn]

Thank you very much for sharing!
Is the loss function you use in the network GIoU or CIoU?

[WongKinYiu]

Provided pre-trained weights are trained by GIoU.
CIoU will get a little bit higher AP.

[kongbia]

Provided pre-trained weights are trained by GIoU.
CIoU will get a little bit higher AP.

Does it need to modify any additional parameters to use CIOU like 'iou_t' in train.py?

[WongKinYiu]

just change GIoU=Ture to CIoU=True.
https://github.com/WongKinYiu/PyTorch_YOLOv4/blob/master/utils/utils.py#L410

[kongbia]

just change GIoU=Ture to CIoU=True.
https://github.com/WongKinYiu/PyTorch_YOLOv4/blob/master/utils/utils.py#L410

Thank you!