Long-Tailed Visual Recognition via Permutation-Invariant Head-to-Tail Feature Fusion

TL;DR

This paper introduces PI-H2T, a versatile method that improves long-tailed visual recognition by enhancing feature representation and classifier bias correction, leading to better tail class recognition.

Contribution

The paper proposes a novel permutation-invariant head-to-tail feature fusion method that improves long-tailed recognition by enhancing feature clustering and transferring semantic information.

Findings

PI-H2T improves tail class accuracy on benchmarks.

The method enhances feature space clustering and class margin formation.

PI-H2T seamlessly integrates with existing models, boosting performance.

Abstract

The imbalanced distribution of long-tailed data presents a significant challenge for deep learning models, causing them to prioritize head classes while neglecting tail classes. Two key factors contributing to low recognition accuracy are the deformed representation space and a biased classifier, stemming from insufficient semantic information in tail classes. To address these issues, we propose permutation-invariant and head-to-tail feature fusion (PI-H2T), a highly adaptable method. PI-H2T enhances the representation space through permutation-invariant representation fusion (PIF), yielding more clustered features and automatic class margins. Additionally, it adjusts the biased classifier by transferring semantic information from head to tail classes via head-to-tail fusion (H2TF), improving tail class diversity. Theoretical analysis and experiments show that PI-H2T optimizes both the representation space and decision boundaries. Its plug-and-play design ensures seamless integration into existing methods, providing a straightforward path to further performance improvements. Extensive experiments on long-tailed benchmarks confirm the effectiveness of PI-H2T.