COCAS
收藏资源简介:
COCAS数据集是由中国科学院深圳先进技术研究院创建的一个大规模行人重识别基准数据集,专注于解决衣物变化下的识别问题。该数据集包含5,266个身份,共计62,382张身体图像,每个身份平均有12张图像,涵盖多种真实场景,如不同的光照和遮挡条件。数据集的构建过程复杂,涉及图像采集、人物与衣物关联等步骤,旨在通过提供同一人物不同衣物的图像,增强模型在实际应用中的识别能力,特别适用于追踪更换衣物的嫌疑人或寻找丢失的儿童和老人。
The COCAS dataset is a large-scale person re-identification benchmark dataset created by the Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, focusing on addressing the recognition problem under clothing variations. It contains 5,266 identities and a total of 62,382 body images, with an average of 12 images per identity, covering various real-world scenarios such as different lighting conditions and occlusion situations. The construction process of the dataset is complex, involving steps like image collection and the association between persons and their clothing. It aims to enhance the recognition capability of models in real-world applications by providing images of the same person wearing different garments, and is particularly suitable for tracking suspects who have changed their clothing or searching for missing children and elderly people.

- 1COCAS: A Large-Scale Clothes Changing Person Dataset for Re-identification中国科学院深圳先进技术研究院 · 2020年



