AlignedReID: Surpassing Human-Level Performance in Person Re-Identification

TL;DR

This paper introduces AlignedReID, a person re-identification method that combines global and local features, achieving surpassing human-level accuracy on major datasets without extra supervision.

Contribution

The novel joint learning approach of global and local features in AlignedReID improves re-identification accuracy beyond previous methods and human performance.

Findings

Achieves 94.4% rank-1 accuracy on Market1501

Achieves 97.8% rank-1 accuracy on CUHK03

Surpasses human-level performance on both datasets

Abstract

In this paper, we propose a novel method called AlignedReID that extracts a global feature which is jointly learned with local features. Global feature learning benefits greatly from local feature learning, which performs an alignment/matching by calculating the shortest path between two sets of local features, without requiring extra supervision. After the joint learning, we only keep the global feature to compute the similarities between images. Our method achieves rank-1 accuracy of 94.4% on Market1501 and 97.8% on CUHK03, outperforming state-of-the-art methods by a large margin. We also evaluate human-level performance and demonstrate that our method is the first to surpass human-level performance on Market1501 and CUHK03, two widely used Person ReID datasets.