question

Hello, may I ask how to use box-level (copy_pastev3.py) in the training process (train_MMD.py)? I see that there is no such part in the training code?

When the dataloader is created, a load_mosaic function is called.

AcroFOD/utils/datasets.py

Line 506 in 53445c4

img, labels = load_mosaic(self, index, copy_paste_flag, cp_type)

The function is exactly the same as the original yolo v5, but they replaced the yolo copy_paste with their own copy_paste function.

Yes. I integrate the copy-paste augmentation into the "load_mosaic" function. The advanced version of yolov5 includes copy_paste augmentation, while the 3.0 version used in this repository does not have such augmentation yet. I am not sure which one can get better results. You can try them :).