This dataset involves 114 different pedestrians and contains 1,824 images, which are taken from 8 directions using two disjoint cameras. For each person, eight images are captured from eight different directions under one camera view, with a standard size of 128×48 pixels. The dataset is also randomly divided into two parts: 57 people are used for training and 57 people are used for testing.

Citation

@article{ma2016orientation,
title={Orientation driven bag of appearances for person re-identification},
author={Ma, Liqian and Liu, Hong and Hu, Liang and Wang, Can and Sun, Qianru},
journal={arXiv preprint arXiv:1605.02464},
year={2016}
}