Resources

⭐ If our datasets and code help your research, a star on GitHub is the best support! ⭐
_{🌟 We are always looking for self-motivated PhD students / RAs / visiting students — see Join Us.}

---

🚁 UAV & Spatial Intelligence

Benchmarks, code and datasets for low-altitude aerial perception: cross-view geo-localization, BEV understanding and aerial reasoning.

🎓 The University-1652 Family

🚀 New Open-Source Releases

Community & Ecosystem

• 🛩️ Host of the UAVM Workshop Series on UAVs in Multimedia @ ACM Multimedia (2023–2026).
• 📚 Awesome Geo-localization: curated papers, datasets and leaderboards.

---

🔍 Person Re-ID & Text-based Retrieval

Datasets

	PAB Dataset ICCV 2025 Highlight [website] We propose a new task, text-based person anomaly search, locating pedestrians engaged in both routine or anomalous activities via text. To enable the training and evaluation of this new task, we construct a large-scale image-text Pedestrian Anomaly Behavior (PAB) benchmark, featuring a broad spectrum of actions, e.g., running, performing, playing soccer, and the corresponding anomalies, e.g., lying, being hit, and falling of the same identity. The training set of PAB comprises 1,013,605 synthesized image-text pairs of both normalities and anomalies, while the test set includes 1,978 real-world image-text pairs.
	MALS Dataset ACM MM 2023 [website] We present a large Multi-Attribute and Language Search dataset for text-based person retrieval, called MALS, and explore the feasibility of performing pre-training on both attribute recognition and image-text matching tasks in one stone. In particular, MALS contains 1, 510, 330 image-text pairs, which is about 37.5× larger than prevailing CUHK-PEDES, and all images are annotated with 27 attributes.
	Market-1501 and DukeMTMC-reID Attribute Datasets [website] We manually annotate attribute labels for two large-scale re-ID datasets, and systematically investigate how person re-ID and attribute recognition benefit from each other.
	3D Market-1501 Dataset [website] You could find the point-cloud format Market-1501 Dataset at https://github.com/layumi/person-reid-3d.
	DG-Market Dataset We provide our generated images and make a large-scale synthetic dataset called DG-Market. This dataset is generated by our DG-Net and consists of 128,307 images (613MB), about 10 times larger than the training set of original Market-1501 (even much more can be generated with DG-Net). It can be used as a source of unlabeled training dataset for semi-supervised learning. You may download the dataset from [Google Drive] (or [Baidu Disk]) password: qxyh).

📦 Legacy Datasets & Tutorials (click to expand)

• HQ-Market Super-resolution Dataset. [website]
• DukeMTMC-reID Dataset. [website] [SoTA]
• DukeMTMC-Pose Dataset. [website]
• UTS Person-reID Tutorial. [website]

---

Awesome Lists

• Awesome Geo-localization
• Awesome Vehicle Retrieval
• Awesome Segmentation Domain Adaptation
• Awesome Fools

---

🎓 For Students

Motivations

• The illustrated guide to a Ph.D.
• 熊辉: 为什么人前进的路总是被自己挡住
• 陈海波: 一名系统研究者的攀登之路
• 勇气与真意——关于围棋的八卦
• 山世光：致联系报考我免试研究生的同学们

How to

• How to research? Jianxiong Xiao
• How to have productive meetings with busy mentors?
• How to start writing papers?
• How to rebuttal? Devi Parikh

---

_{Last updated: July 2026. If you build something cool on top of our benchmarks, we are happy to feature it — drop us an email!}