Supplementary: Unleashing Potential of Unsupervised Pre-Training with Intra-Identity Regularization for Person Re-Identification

Zizheng Yang,Xin Jin,Kecheng Zheng,Feng Zhao
2022-01-01
Abstract:CUHK03 [10] contains 13,164 images of 1,360 pedestrians. Each identity is observed by 2 cameras. Note that CUHK03 offers both hand-labeled and DPM-detected bounding boxes, and the former is adopted in this paper. Market1501 [22] contains 32,668 person images of 1,501 identities captured by 6 cameras. The training set consists of 12,936 images of 751 identities, the query set consists of 3,368 images, and the gallery set consists of 19,732 images of 750 identities. PersonX [14] is a large-scale data synthesis engine, which contains 1,266 manually designed identities and editable visual variables. Each identity is captured by 6 cameras. MSMT17 [18] contains of 126,441 images of 4,101 identities captured by 15 cameras. The training set consists of 30,248 person images of 1,041 identities, the query set consists of 11,659 images, and the gallery consists of 82,161 images of 3,060 identities.
What problem does this paper attempt to address?