NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:81
Title:FreeAnchor: Learning to Match Anchors for Visual Object Detection

The paper presents a better loss function for anchor-based detection methods by matching anchors to GT boxes in a differentiable manner. Results are convincing. Three reviewers recommend acceptance after a convincing rebuttal. The final decision is to accept.