Did you use head loss same as the author?

[Edwardmark]

The paper downsample the input image region correcbonding to head anchor by 2 to match the face gt, but I cannot find out where you did this operation in your code. Did I miss something? Please comment @Goingqs.

[Edwardmark]

I cannot find similar code in your implementation, as https://github.com/PaddlePaddle/models/blob/develop/fluid/face_detection/reader.py#L205

[Edwardmark]

@Goingqs Please comment, how you match anchor to head GT(which is expand gt of face gt).