Receptive Multi-granularity Representation for Person Re-Identification

TL;DR

This paper introduces a receptive multi-granularity learning method for person re-identification that improves local feature consistency and robustness by adaptively partitioning intermediate representations, achieving state-of-the-art results.

Contribution

It proposes a novel local partitioning approach on intermediate features with significance-balanced pooling and data augmentation, enhancing local detail consistency in person re-ID.

Findings

Achieves 96.2% Rank-1 accuracy on Market-1501

Outperforms existing methods in intra- and cross-dataset evaluations

Provides comprehensive feature representation without increased model size

Abstract

A key for person re-identification is achieving consistent local details for discriminative representation across variable environments. Current stripe-based feature learning approaches have delivered impressive accuracy, but do not make a proper trade-off between diversity, locality, and robustness, which easily suffers from part semantic inconsistency for the conflict between rigid partition and misalignment. This paper proposes a receptive multi-granularity learning approach to facilitate stripe-based feature learning. This approach performs local partition on the intermediate representations to operate receptive region ranges, rather than current approaches on input images or output features, thus can enhance the representation of locality while remaining proper local association. Toward this end, the local partitions are adaptively pooled by using significance-balanced activations for uniform stripes. Random shifting augmentation is further introduced for a higher variance of person appearing regions within bounding boxes to ease misalignment. By two-branch network architecture, different scales of discriminative identity representation can be learned. In this way, our model can provide a more comprehensive and efficient feature representation without larger model storage costs. Extensive experiments on intra-dataset and cross-dataset evaluations demonstrate the effectiveness of the proposed approach. Especially, our approach achieves a state-of-the-art accuracy of 96.2%@Rank-1 or 90.0%@mAP on the challenging Market-1501 benchmark.