Frequency Domain Modality-invariant Feature Learning for Visible-infrared Person Re-Identification

TL;DR

This paper introduces FDMNet, a frequency domain framework that reduces modality discrepancy in visible-infrared person re-identification by focusing on amplitude components, achieving superior results on standard benchmarks.

Contribution

The paper reveals amplitude differences as the main cause of modality discrepancy and proposes FDMNet with novel modules to address this in the frequency domain.

Findings

FDMNet outperforms state-of-the-art methods on SYSU-MM01 and RegDB benchmarks.

The amplitude component is identified as the primary factor in modality discrepancy.

Frequency domain approach effectively reduces cross-modality differences.

Abstract

Visible-infrared person re-identification (VI-ReID) is challenging due to the significant cross-modality discrepancies between visible and infrared images. While existing methods have focused on designing complex network architectures or using metric learning constraints to learn modality-invariant features, they often overlook which specific component of the image causes the modality discrepancy problem. In this paper, we first reveal that the difference in the amplitude component of visible and infrared images is the primary factor that causes the modality discrepancy and further propose a novel Frequency Domain modality-invariant feature learning framework (FDMNet) to reduce modality discrepancy from the frequency domain perspective. Our framework introduces two novel modules, namely the Instance-Adaptive Amplitude Filter (IAF) module and the Phrase-Preserving Normalization (PPNorm) module, to enhance the modality-invariant amplitude component and suppress the modality-specific component at both the image- and feature-levels. Extensive experimental results on two standard benchmarks, SYSU-MM01 and RegDB, demonstrate the superior performance of our FDMNet against state-of-the-art methods.