Learning Rotation-Equivariant Features for Visual Correspondence

TL;DR

This paper presents a self-supervised framework using group-equivariant CNNs to learn rotation-invariant local features for image correspondence, improving robustness to rotations without data augmentation.

Contribution

Introduces a novel self-supervised learning method with group-equivariant CNNs and a group aligning technique for rotation-invariant feature extraction.

Findings

Achieves state-of-the-art rotation-invariant matching accuracy

Demonstrates robustness to geometric and photometric variations

Effective transfer to keypoint matching and pose estimation tasks

Abstract

Extracting discriminative local features that are invariant to imaging variations is an integral part of establishing correspondences between images. In this work, we introduce a self-supervised learning framework to extract discriminative rotation-invariant descriptors using group-equivariant CNNs. Thanks to employing group-equivariant CNNs, our method effectively learns to obtain rotation-equivariant features and their orientations explicitly, without having to perform sophisticated data augmentations. The resultant features and their orientations are further processed by group aligning, a novel invariant mapping technique that shifts the group-equivariant features by their orientations along the group dimension. Our group aligning technique achieves rotation-invariance without any collapse of the group dimension and thus eschews loss of discriminability. The proposed method is trained end-to-end in a self-supervised manner, where we use an orientation alignment loss for the orientation estimation and a contrastive descriptor loss for robust local descriptors to geometric/photometric variations. Our method demonstrates state-of-the-art matching accuracy among existing rotation-invariant descriptors under varying rotation and also shows competitive results when transferred to the task of keypoint matching and camera pose estimation.