Repeatability Is Not Enough: Learning Affine Regions via Discriminability

TL;DR

This paper introduces a new method for learning affine regions that emphasizes descriptor discriminability over repeatability, using a novel loss function, resulting in improved image retrieval and stereo matching.

Contribution

It proposes a hard negative-constant loss for learning affine regions, outperforming existing methods without requiring precisely aligned patches.

Findings

Affine shape estimator (AffNet) outperforms state-of-the-art in image retrieval.

The proposed training method does not need geometrically aligned patches.

The method improves wide baseline stereo matching.

Abstract

A method for learning local affine-covariant regions is presented. We show that maximizing geometric repeatability does not lead to local regions, a.k.a features,that are reliably matched and this necessitates descriptor-based learning. We explore factors that influence such learning and registration: the loss function, descriptor type, geometric parametrization and the trade-off between matchability and geometric accuracy and propose a novel hard negative-constant loss function for learning of affine regions. The affine shape estimator -- AffNet -- trained with the hard negative-constant loss outperforms the state-of-the-art in bag-of-words image retrieval and wide baseline stereo. The proposed training process does not require precisely geometrically aligned patches.The source codes and trained weights are available at https://github.com/ducha-aiki/affnet