NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:6717
Title:R2D2: Reliable and Repeatable Detector and Descriptor

The work tackles the problem of local feature detection and description, and proposes a novel loss function for learning both in an end-to-end pipeline. The reviewers are uniformly aligned in favor of acceptance. The (very limited) initial concerns were addressed by the rebuttal.