search

fundamentalvision / Deformable-DETR

Deformable DETR: Deformable Transformers for End-to-End Object Detection.
Apache License 2.0
3.22k stars 520 forks source link

Why is the classification cost in matcher formulated as the positive minus negative? #200

Closed Godricly closed 1 year ago

Godricly commented 1 year ago

Hi, I'm confused with this part. The computation here is different from the implementation in sigmoid_focal_loss computation. What is the purpose here to use minus sign? https://github.com/fundamentalvision/Deformable-DETR/blob/11169a60c33333af00a4849f1808023eba96a931/models/matcher.py#L81

Godricly commented 1 year ago

found an explaination in https://arxiv.org/pdf/2304.04742.pdf.