DetReIDX: A Stress-Test Dataset for Real-World UAV-Based Person Recognition (IEEE T-BIOM 2026) PDF Project Page GitHub Leaderboard
- 2026.03.06 — 1st Workshop on VReID-XFD at WACV 2026, Tucson, Arizona
- 2026 — Paper accepted at IEEE Transactions on Biometrics, Behavior, and Identity Science (T-BIOM)
- 2025.05 — Preprint available on arXiv
- Build project page
- Person Re-ID split released
- Person Detection split — Coming Soon
- Video ReID split — Coming Soon
DetReIDX: Long-range person identification from UAV and ground cameras. Click to watch.
Figure 1: Comparison of DetReIDX with existing datasets across altitude coverage, annotation richness, and task support.
- 509 unique identities | 13M+ bounding box annotations | 7 international collection sites
- 18 UAV viewpoints per subject — altitudes 5m to 120m, distances 10m to 120m, pitch angles 30°/60°/90°
- Clothing variation across sessions — tests long-term identity persistence
- 16 soft biometric attributes per frame — demographics, appearance, clothing, action
- 4 tasks: Person Detection · Re-ID · Multi-View Tracking · Action Recognition
| Setting | Environment | Altitude | Distance | Data Type |
|---|---|---|---|---|
| Indoor | University Labs | Ground level | Close range | Profile images, gait videos |
| Outdoor | University Campuses | 5–120m | 10–120m | Multi-view videos, action clips |
| Indoor Lab Setup | Outdoor UAV Setup |
|---|---|
![]() |
![]() |
Teaser: intro2.mp4
Outdoor UAV Examples:
| Site | Video |
|---|---|
| Istanbul Medipol University | Watch |
| JNNCE, India | Watch |
| SRMIST, India | Watch |
| SRTMUN, India | Watch |
| University of Luanda, Angola | Watch |
| NMDC | Watch |
Indoor Reference Session: indoor_sample_video.mp4
| # | Attribute | Values |
|---|---|---|
| 1 | Gender | Male, Female, Unknown |
| 2 | Age | 0–11, 12–17, 18–24, 25–34, 35–44, 45–54, 55–64, >65, Unknown |
| 3 | Height | Child, Short, Medium, Tall, Unknown |
| 4 | Body Volume | Thin, Medium, Fat, Unknown |
| 5 | Ethnicity | White, Black, Asian, Indian, Unknown |
| 6 | Hair Color | Black, Brown, White, Red, Gray, Occluded, Unknown |
| 7 | Hairstyle | Bald, Short, Medium, Long, Ponytail, Unknown |
| 8 | Beard | Yes, No, Unknown |
| 9 | Moustache | Yes, No, Unknown |
| 10 | Glasses | Normal, Sun, No, Unknown |
| 11 | Head Accessories | Hat, Scarf, Necklace, Cannot see, Unknown |
| 12 | Upper Body Clothing | T-Shirt, Blouse, Sweater, Coat, Dress, Uniform, Shirt, Suit, Hoodie, Cardigan, Unknown |
| 13 | Lower Body Clothing | Jeans, Leggings, Pants, Shorts, Skirt, Dress, Uniform, Suit, Unknown |
| 14 | Feet | Sport Shoe, Classic Shoe, High Heels, Boots, Sandal, Nothing, Unknown |
| 15 | Accessories | Bag, Backpack, Rolling Bag, Umbrella, Sport Bag, Market Bag, Nothing, Unknown |
| 16 | Action | Walking, Running, Standing, Sitting, Cycling, Exercising, Talking on Phone, and more |
Full encoding details: soft_bio_readme.md
Files are distributed as password-protected ZIP archives. After downloading, email with your name, institution, and intended use to receive the password.
📧 n1947@ubi.pt or kailas.srt@gmail.com — Subject: "DetReIDX Password Request"
| Split | Status | Link |
|---|---|---|
| Person Re-ID (still + video, with attribute labels) | ✅ Available | Google Drive |
| Person Detection (bounding box annotations) | 🕐 Coming Soon | — |
| Video ReID (multi-session tracklets) | 🕐 Coming Soon | — |
📄 Download License Agreement (PDF)
DetReIDX/
├── train/
│ └── <PID>/
│ └── <tracklet>/
│ └── *.jpg
├── query/
│ └── <PID>/
│ └── <tracklet>/
│ └── *.jpg
└── gallery/
└── <PID>/
└── <tracklet>/
└── *.jpg
| Protocol | Task | Metric |
|---|---|---|
| Short-Term ReID | Match within same session | mAP, CMC Rank-1/5/10 |
| Long-Term ReID | Match across sessions (clothing change) | mAP, CMC Rank-1/5/10 |
| Cross-View ReID | Aerial ↔ Ground matching | mAP, CMC Rank-1/5/10 |
| Detection | Pedestrian bounding box prediction | mAP@0.5, mAP@0.75 |
| Tracking | Multi-view person tracking | MOTA, IDF1 |
Full experimental results (YOLOv8, DDOD, Grid-RCNN, PersonViT, SeCap, CLIP-ReID) in the paper.
We acknowledge and give credit to the following universities for their contributions: Istanbul Medipol University · J.N.N College of Engineering · SRM Institute of Science and Technology · Swami Ramanand Teerth Marathwada University Nanded · Universidade Beira Interior · Universidade de Luanda (A–Z)
@article{hambarde2026detreidx,
title = {DetReIDX: A Stress-Test Dataset for Real-World UAV-Based Person Recognition},
author = {Hambarde, Kailash A and Mbongo, Nzakiese and Kumar, MP Pavan and Mekewad, Satish
and Fernandes, Carolina and Silahtaro{\u{g}}lu, G{\"o}khan and Nithya, Alice
and Wasnik, Pawan and Rashidunnabi, MD and Samale, Pranita and others},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2026},
publisher = {IEEE}
}Released for academic and non-commercial use only. By downloading or using this dataset, you agree to the terms on the project website.



