DeepChange: A Long-Term Person Re-Identification Benchmark with Clothes Change ICCV 2023

This is the official project page of the paper "DeepChange: A Long-Term Person Re-Identification Benchmark with Clothes Change" ICCV 2023.



Authors: XU Peng and ZHU Xiatian



We contribute a large, realistic long-term person re-identification benchmark. It consists of 178K bounding boxes from 1.1K person identities, collected and constructed over a course of 12 months. Unique characteristics of this dataset include: (1) Natural/native personal appearance (e.g., clothes and hair style) variations: The clothes-change and dressing styles are highly diverse, with the reappearing gap in time ranging from minutes, hours, and days to weeks, months, seasons, and years. (2) Diverse walks of life: Persons across a wide range of ages and professions appear in different weather conditions (e.g., sunny, cloudy, windy, rainy, snowy, extremely cold) and events (e.g., working, leisure, daily activities). (3) Rich camera setups: The raw videos were recorded by 17 outdoor security cameras with various resolutions operating in a real-world surveillance system for a wide and dense block. (4) Largest scale: It covers the largest number of (17) cameras, (1121) identities, and (178407) bounding boxes, as compared to alternative datasets.

Figure A. Images of a random person identity from DeepChange with clothes change, as well as variation in behavior, pose, hair, illumination, weather, background, etc.

Figure B. Image samples of random identities in DeepChange. Identities from top left to bottom right: an aunt (bbox#1-#19), an office lady (bbox#20-#27), a pupil (bbox#28-#36), a newspaper delivery (bbox#37-#41), an older aunt (bbox#42-#49), a worker (bbox#50-#51), a nun (bbox#52-#53), a Muslim man (bbox#54-#56), a chef (bbox#57-#58), a disabled person (bbox#59-#60), a dustman (bbox#61-#70).

Figure C. Images collected in snow (top two lines, bbox#1-#20) and rain (bottom line, bbox#21-#30) weather from DeepChange.

Figure D. Image pairs of the specific persons with simultaneous clothes and hair style changes in the DeepChange. For each identity, only two cases are selected randomly.

Terms and conditions to apply

• This dataset can be ONLY used for non-commercial research purposes.
• Users are NOT allowed to reproduce, duplicate, copy, sell, trade, resell or exploit any portion of the images and any portion of derived data for any commercial purposes.
• Users are NOT allowed to further publish or distribute any portion of this dataset.
• Users MUST block the faces of any persons from this dataset when using images for any dissemination purposes, e.g., academic papers / slides / posters, presentations, personal websites, personal blogs.
• ALL the rights to terminate the access are reserved.

Download

Users need to download and sign our DeepChange Dataset Access Agreement, and send it to XU Peng (peng.xu [AT] aliyun dOt com) with a title "DeepChange Dataset Access Agreement". We will verify the requests and contact users with the downloading links and other associate documents. Users should use your institutional email address to contact us. We will reply within 24 hours, and the download link is also valid for 24 hours, so after sending this request, please watch your email.
请大家用学校的邮箱发送申请表格。请勿使用163或qq等邮箱发送申请。请字迹工整地且完整地填写申请表格，可直接填写中文，Email 2里请填写一个可以永久联系的私人邮箱。我们通常会在24小时内回复，授权后下载链接有效期是24小时，所以发送申请后，请关注您的邮箱并及时下载。

Software tools

We will release some codes and tools for this dataset very soon.

Citations

If you would like to use this dataset, please cite the following papers:

@inproceedings{xu2023deepchange,
  title={DeepChange: A Long-Term Person Re-Identification Benchmark with Clothes Change},
  author={Xu, Peng and Zhu, Xiatian},
  booktitle={Proceedings of the IEEE international conference on computer vision (ICCV)},
  year={2023}
}

Contact

Please email XU Peng for any questions.
Email: peng.xu [AT] aliyun dOt com