Rethinking Illumination for Person Re-Identification: A Unified View

Suncheng Xiang, Guanjie You, Leqi Li, Mengyuan Guan, Ting Liu, Dahong Qian, Yuzhuo Fu; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2022, pp. 4731-4739

As a fundamental problem in video surveillance, person re-identification (re-ID) contributes a lot to the development of modern metro city. Recently, learning from synthetic data on re-ID task, which benefits from the popularity of synthetic data engine, has achieved remarkable performance in both supervised and unsupervised manner. However, previous researches mainly lay emphasis on employing synthetic data to achieve the state-of-the-art performance with a strong backbone, while neglects to perform quantitative studies on how visual factors affect re-ID system. To facilitate the research in this field, firstly, we manually construct a large-scale synthetic dataset named SynPerson, which has diversified human characters and distinguished attributes with accurate annotations. Secondly, we quantitatively analyze the influence of illumination on re-ID system. To our best knowledge, this is the first attempt to explicitly dissect person re-ID from the aspect of illumination on synthetic dataset. Comprehensive experiments help us have a deeper understanding of the fundamental problems in person re-ID. Furthermore, we will release SynPerson to the community, as part of efforts to alleviate the shortage of large-scale pedestrian dataset of future works.