DISK: Learning local features with policy gradient

Michał Tyszkiewicz, Pascal Fua, Eduard Trulls

Spotlight presentation: Orals & Spotlights Track 07: Vision Applications
on 2020-12-08T07:00:00-08:00 - 2020-12-08T07:10:00-08:00
Poster Session 2 (more posters)
on 2020-12-08T09:00:00-08:00 - 2020-12-08T11:00:00-08:00
GatherTown: Vision ( Town A2 - Spot D2 )
Join GatherTown
Only iff poster is crowded, join Zoom . Authors have to start the Zoom call from their Profile page / Presentation History.
Abstract: Local feature frameworks are difficult to learn in an end-to-end fashion due to the discreteness inherent to the selection and matching of sparse keypoints. We introduce DISK (DIScrete Keypoints), a novel method that overcomes these obstacles by leveraging principles from Reinforcement Learning (RL), optimizing end-to-end for a high number of correct feature matches. Our simple yet expressive probabilistic model lets us keep the training and inference regimes close, while maintaining good enough convergence properties to reliably train from scratch. Our features can be extracted very densely while remaining discriminative, challenging commonly held assumptions about what constitutes a good keypoint, as showcased in Fig. 1, and deliver state-of-the-art results on three public benchmarks.

Preview Video and Chat

Chat is not available.