Gradients for bounding box refinement

vacancy / PreciseRoIPooling

Precise RoI Pooling with coordinate gradient support, proposed in the paper "Acquisition of Localization Confidence for Accurate Object Detection" (https://arxiv.org/abs/1807.11590).

MIT License

770 stars 152 forks source link

Gradients for bounding box refinement #41

Closed AlexanderHustinx closed 4 years ago

AlexanderHustinx commented 4 years ago

Dear authors,

I have a question regarding Algorithm 2: Optimization-based bounding box refinement. In Algorithm 2 you set **grad** <- \nabla_{b_j} IoU(PrPool(F, b_j)) Later you also mention: "... the proposed PreciseRoI Pooling layer enables the computation of the gradient of IoU w.r.t. bounding box coordinates."

How do you get these gradient values, grad (in PyTorch)?

Thanks in advance!

vacancy commented 4 years ago

Estimate the IoU using the proposed IoU prediction branch, denote the output tensor as iou.
Run iou.backward()
The gradients are b_j.grad.