search

JialeCao001 / D2Det

D2Det: Towards High Quality Object Detection and Instance Segmentation (CVPR2020)
https://openaccess.thecvf.com/content_CVPR_2020/papers/Cao_D2Det_Towards_High_Quality_Object_Detection_and_Instance_Segmentation_CVPR_2020_paper.pdf
MIT License
297 stars 86 forks source link

Question about Discriminative RoI Pooling #25

Closed cookchicken closed 3 years ago

cookchicken commented 3 years ago

Firstly, thank you for such a greate work. But I have a little problem when I reading your code. In the paper I know that the Discriminative RoI pooling create the Weighted RoI feature with F~ =W(F)⊙ F. But In your code the Weighted RoI feature is created by return F.avg_pool2d(x * weight, kernel, stride, padding) / F.avg_pool2d(weight, kernel, stride, padding). I don't know why do it divide by F.avg_pool2d(weight, kernel, stride, padding)

JialeCao001 commented 3 years ago

@cookchicken Thanks for your interest. It is used for normalization.

cookchicken commented 3 years ago

@cookchicken Thanks for your interest. It is used for normalization.

@cookchicken Thanks for your interest. It is used for normalization.