Learning with Label Noise for Image Retrieval by Selecting Interactions

TL;DR

This paper introduces T-SINT, a noise-resistant method for image retrieval that identifies and selects correct interactions in the presence of noisy labels, improving retrieval performance under high noise conditions.

Contribution

The paper proposes T-SINT, a novel teacher-based approach that effectively handles noisy labels in image retrieval by selecting reliable interactions, which was less explored before.

Findings

Outperforms state-of-the-art methods on benchmark datasets with synthetic noise

Effective in high noise rate scenarios

Demonstrates robustness with realistic noise conditions

Abstract

Learning with noisy labels is an active research area for image classification. However, the effect of noisy labels on image retrieval has been less studied. In this work, we propose a noise-resistant method for image retrieval named Teacher-based Selection of Interactions, T-SINT, which identifies noisy interactions, ie. elements in the distance matrix, and selects correct positive and negative interactions to be considered in the retrieval loss by using a teacher-based training setup which contributes to the stability. As a result, it consistently outperforms state-of-the-art methods on high noise rates across benchmark datasets with synthetic noise and more realistic noise.