Abstract

While most steps in the modern object detection methods are learnable, theregion feature extraction step remains largely hand-crafted, featured by RoIpooling methods. This work proposes a general viewpoint that unifies existingregion feature extraction methods and a novel method that is end-to-endlearnable. The proposed method removes most heuristic choices and outperformsits RoI pooling counterparts. It moves further towards fully learnable objectdetection.

 

Quick Read (beta)