Cascade Transformers for End-to-End Person Search

TL;DR

This paper introduces COAT, a three-stage cascade transformer model that progressively refines person detection and re-identification, effectively handling occlusions and variations to achieve state-of-the-art results in person search.

Contribution

The paper presents a novel cascade occluded attention transformer that refines person search through multiple stages, incorporating occluded attention to improve robustness against occlusions and pose variations.

Findings

Achieves state-of-the-art performance on benchmark datasets.

Effectively handles occlusions and pose variations.

Demonstrates the benefit of multi-stage refinement in person search.

Abstract

The goal of person search is to localize a target person from a gallery set of scene images, which is extremely challenging due to large scale variations, pose/viewpoint changes, and occlusions. In this paper, we propose the Cascade Occluded Attention Transformer (COAT) for end-to-end person search. Our three-stage cascade design focuses on detecting people in the first stage, while later stages simultaneously and progressively refine the representation for person detection and re-identification. At each stage the occluded attention transformer applies tighter intersection over union thresholds, forcing the network to learn coarse-to-fine pose/scale invariant features. Meanwhile, we calculate each detection's occluded attention to differentiate a person's tokens from other people or the background. In this way, we simulate the effect of other objects occluding a person of interest at the token-level. Through comprehensive experiments, we demonstrate the benefits of our method by achieving state-of-the-art performance on two benchmark datasets.