Harmonious Attention Network for Person Re-Identification

TL;DR

This paper introduces HA-CNN, a novel model that jointly learns attention mechanisms and feature representations to improve person re-identification in uncontrolled, misaligned images, outperforming existing methods on multiple benchmarks.

Contribution

The paper proposes a new Harmonious Attention CNN that jointly optimizes pixel and regional attention with feature learning for better re-id in challenging scenarios.

Findings

HA-CNN outperforms state-of-the-art methods on CUHK03, Market-1501, and DukeMTMC-ReID datasets.

Joint attention and feature learning improve re-id accuracy in misaligned images.

Extensive experiments validate the effectiveness of the proposed approach.

Abstract

Existing person re-identification (re-id) methods either assume the availability of well-aligned person bounding box images as model input or rely on constrained attention selection mechanisms to calibrate misaligned images. They are therefore sub-optimal for re-id matching in arbitrarily aligned person images potentially with large human pose variations and unconstrained auto-detection errors. In this work, we show the advantages of jointly learning attention selection and feature representation in a Convolutional Neural Network (CNN) by maximising the complementary information of different levels of visual attention subject to re-id discriminative learning constraints. Specifically, we formulate a novel Harmonious Attention CNN (HA-CNN) model for joint learning of soft pixel attention and hard regional attention along with simultaneous optimisation of feature representations, dedicated to optimise person re-id in uncontrolled (misaligned) images. Extensive comparative evaluations validate the superiority of this new HA-CNN model for person re-id over a wide variety of state-of-the-art methods on three large-scale benchmarks including CUHK03, Market-1501, and DukeMTMC-ReID.