How to scale ground-truth to the feature layers?

[aidevmin]

@twangnh
In the paper you wrote The very simple setting (0.25×-G) of directly scaling ground truth boxes with same stride on the feature layer and applying imitation on those areas gives much less gain than the proposed method. I am interested to get GT-mask in the feature map.
Could you give any suggestion?

With YOLO (v5, v6, v7, v8) model, we have 3 feature maps. How can we know map one ground-truth box to which feature map?
Thanks a lot.

[twangnh]

please refer to https://github.com/wonbeomjang/yolov5-knowledge-distillation for a example

[aidevmin]

@twangnh
Thanks for response.
With this repo https://github.com/wonbeomjang/yolov5-knowledge-distillation, we can not open issue. I have read this repo true, but it does not mention about GT-mask.

[twangnh]

could you run the code to find how it generate GT- mask