Open RichardcLee opened 2 years ago
and Why do you use this function:clamp(min=1e-6) ?
Hi, this repo is implemented using the framework of MMDetection which you may need to get familiar with.
VFNet mainly involves a new loss, Varifocal Loss, and a detector head, vfnet_head. Most of the details are in these two scripts and understanding them is almost enough to reimplement VFNet.
clamp(min=1e-6) ?
This is to avoid the calculated IoU to be 0.
I want to reimplement VFNet by detectron2. Can you describe the technical details of implementing it in detail? please.