An Adaptive Part-Based Model For Person Re-Identification

Abstract

Existing part-based models for person Re-IDentification (Re-ID) usually suffer from part-misalignment problem caused by uniform partition of feature maps. The performances of part-based model are highly dependent on the semantically-aligned parts of the query and gallery images. However, misalignments occur very commonly in person Re-ID tasks due to the variations of viewpoints and object distances. To address the part-misalignment problem and learn a more discriminative embedding for person Re-ID, we propose a novel Adaptive Part-based Model (APM), which adaptively partition the extracted feature maps by a Partition-Aware module to learn an embedding. The proposed adaptive partition method is very robust to the variations of the pedestrian scale and effective in resolving the part-misalignment problem. Experimental results on three commonly used datasets, including Market-1501, DukeMTMC-reID and CUHK03, clearly demonstrate that the proposed method achieves the state-of-the-art performance.