From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization

TL;DR

This work tackles noise in person re-identification by proposing a training-free feature centralization framework. It leverages the central tendency of same-ID features around identity centers, justified by a multivariate normal distribution, and combines Identity-Guided Pedestrian Generation (IPG) with Neighbor Feature Centralization (NFC) to centralize representations without ReID training. IPG uses diffusion-based synthesis guided by identity features and selected representative poses to produce diverse, identity-consistent images, while NFC exploits neighborhood structure to further refine features. The approach achieves state-of-the-art results across standard, cross-modality, and occluded ReID tasks, including strong performance with ImageNet-pretrained models and without re-ranking, demonstrating practical effectiveness for deployment in surveillance contexts.

Abstract

Person re-identification (ReID) aims to extract accurate identity representation features. However, during feature extraction, individual samples are inevitably affected by noise (background, occlusions, and model limitations). Considering that features from the same identity follow a normal distribution around identity centers after training, we propose a Training-Free Feature Centralization ReID framework (Pose2ID) by aggregating the same identity features to reduce individual noise and enhance the stability of identity representation, which preserves the feature's original distribution for following strategies such as re-ranking. Specifically, to obtain samples of the same identity, we introduce two components: Identity-Guided Pedestrian Generation: by leveraging identity features to guide the generation process, we obtain high-quality images with diverse poses, ensuring identity consistency even in complex scenarios such as infrared, and occlusion. Neighbor Feature Centralization: it explores each sample's potential positive samples from its neighborhood. Experiments demonstrate that our generative model exhibits strong generalization capabilities and maintains high identity consistency. With the Feature Centralization framework, we achieve impressive performance even with an ImageNet pre-trained model without ReID training, reaching mAP/Rank-1 of 52.81/78.92 on Market1501. Moreover, our method sets new state-of-the-art results across standard, cross-modality, and occluded ReID tasks, showcasing strong adaptability.

Figures (11)

• Figure 1: Visualization of our Identity-Guided Pedestrian Generation model with 8 representative poses on three datasets.
• Figure 2: The real feature distribution of images from the same ID extracted by TransReID, and the main idea of our work is to make features closer/centralized to the ID center.
• Figure 3: Overview of the proposed Feature Centralization framework for the ReID task.
• Figure 4: t-SNE visualization of 10 IDs feature distribution with and without our method on ImageNet pre-trained weights.
• Figure 5: The effects with and without the IFR module were visualized with five different poses randomly selected for each reference picture.