DisasterM3: A Remote Sensing Vision-Language Dataset for Disaster Damage Assessment and Response

Junjue Wang*, Weihao Xuan*, Heli Qi, Zhihao Liu, Kunyi Liu, Yuhan Wu, Hongruixuan Chen, Jian Song

Junshi Xia, Zhuo Zheng, Naoto Yokoya†

* Equal Contributions † Corresponding Author

[Paper], [Dataset]

Highlights

DisasterM3 includes 26,988 bi-temporal satellite images and 123k instruction pairs across 5 continents, with three characteristics:

1. Multi-hazard: 36 historical disaster events with significant impacts, which are categorized into 10 common natural and man-made disasters
2. Multi-sensor: Extreme weather during disasters often hinders optical sensor imaging, making it necessary to combine Synthetic Aperture Radar (SAR) imagery for post-disaster scenes
3. Multi-task: 9 disaster-related visual perception and reasoning tasks, harnessing the full potential of VLM's reasoning ability

News

• 2026/09/30, 🔥🔥The second version of our disaster benchmark (DORA) got accepted by NeurIPS 2025 as an Oral paper!🔥🔥
• 2025/10/23, We released the DisasterM3 instruct set.
• 2025/10/17, We released the DisasterM3 benchmark set.
• 2025/09/22, We are preparing the dataset and code.
• 2025/09/22, Our paper got accepted by NeurIPS 2025.

Benchmark

Please run this code for benchmarking the DisasterM3 dataset. Two examples: Qwen2.5 VL:

python disaster_m3/pyscripts/run_vllm.py --model_id Qwen/Qwen2.5-VL-7B-Instruct --subset bearing_body

InternVL3:

python disaster_m3/pyscripts/run_vllm.py --model_id OpenGVLab/InternVL3-78B --subset report

Citation

If you use DisasterM3 in your research, please cite our following papers.

@inproceedings{wang2026dora,
  title={Can LLM Agents Respond to Disasters? Benchmarking Heterogeneous Geospatial Reasoning in Emergency Operations},
  author={Wang, Junjue and Xuan, Weihao and Qi, Heli and Dai, Pengyu and Liu, Kunyi and Chen, Hongruixuan and Zheng, Zhuo and Xia, Junshi and Ermon, Stefano and Yokoya, Naoto},
  booktitle={Proceedings of the Neural Information Processing Systems},
  year={2026}
}

@inproceedings{wang2025disasterm3,
  title={DisasterM3: A Remote Sensing Vision-Language Dataset for Disaster Damage Assessment and Response},
  author={Wang, Junjue and Xuan, Weihao and Qi, Heli and Liu, Zhihao and Liu, Kunyi and Wu, Yuhan and Chen, Hongruixuan and Song, Jian and Xia, Junshi and Zheng, Zhuo and Yokoya, Naoto},
  booktitle={Proceedings of the Neural Information Processing Systems},
  year={2025}
}

Acknowledgments

This dataset builds upon the following excellent open datasets:

• xBD dataset by Ritwik Gupta

  • Paper
  • Dataset
  • License: [CC BY-NC-SA 4.0]

• BRIGHT dataset by Hongruixuan Chen

  • Repository
  • License: [CC BY-NC 4.0]

License

All images and their associated annotations in DisasterM3 can be used for academic purposes only, but any commercial use is prohibited.

Star History