README.md

---
license: mit
task_categories:
- object-detection
- image-segmentation
- robotics
language:
- en
tags:
- code
size_categories:
- 100K<n<1M
---

# Griffin: Aerial-Ground Cooperative Detection and Tracking Dataset and Benchmark

<h3 align="center">
    <a href="https://arxiv.org/abs/2503.06983">📄 arXiv</a> |
    <a href="https://github.com/wang-jh18-SVM/Griffin">🐙 GitHub</a> |
    <a href="https://pan.baidu.com/s/1NDgsuHB-QPRiROV73NRU5g?pwd=u3cm">💾 Baidu Netdisk</a> |
    <a href="https://huggingface.co/datasets/wjh-svm/Griffin">🤗 Hugging Face</a>
</h3>

**Griffin** is the pioneering publicly available dataset for aerial-ground cooperative 3D perception. Built using CARLA-AirSim co-simulation, it features over 200 dynamic scenes—totaling more than 30,000 frames and 270,000 images. With instance-aware occlusion quantification, variable UAV altitudes (20–60 meters), and realistic drone dynamics under diverse conditions, Griffin sets a new standard for cooperative perception research.