Open IssacCyj opened 2 years ago
The output shape of the classifier is 1231: 1230 (category-specific task) + 1 (category-agnostic task).
Does eqlv2 work without the category-agnostic task? I'm trying to implement it on yolov5...
Yes. BTW, if you want to use focal loss, you can use equalized focal loss
Thanks for this great repo.
It mentioned in the paper that you also used category-agnostic task. Is it implemented in this repo? I didn't find it but I was able to reproduce the results in the paper.