Spherical Feature Transform for Deep Metric Learning

TL;DR

This paper introduces a spherical feature transform for deep metric learning that performs class-specific rotations on a hypersphere, improving data augmentation and achieving state-of-the-art results.

Contribution

It proposes a novel spherical feature transform that relaxes covariance assumptions and uses rotations, enhancing deep metric learning performance.

Findings

Achieves consistent performance improvements across benchmarks

Outperforms existing methods with state-of-the-art results

Provides a simple, effective training method

Abstract

Data augmentation in feature space is effective to increase data diversity. Previous methods assume that different classes have the same covariance in their feature distributions. Thus, feature transform between different classes is performed via translation. However, this approach is no longer valid for recent deep metric learning scenarios, where feature normalization is widely adopted and all features lie on a hypersphere. This work proposes a novel spherical feature transform approach. It relaxes the assumption of identical covariance between classes to an assumption of similar covariances of different classes on a hypersphere. Consequently, the feature transform is performed by a rotation that respects the spherical data distributions. We provide a simple and effective training method, and in depth analysis on the relation between the two different transforms. Comprehensive experiments on various deep metric learning benchmarks and different baselines verify that our method achieves consistent performance improvement and state-of-the-art results.