O3N: Omnidirectional Open-Vocabulary Occupancy Prediction

Mengfei Duan^1*, Hao Shi^2*, Fei Teng¹, Guoqiang Zhao¹, Yuheng Zhang¹, Zhiyong Li^1†, Kailun Yang^1†
1Hunan University, ²Zhejiang University

O3N [PDF]

O3N is the first purely visual, end-to-end framework for omnidirectional open-vocabulary occupancy prediction. It enables autonomous agents and embodied AI systems to reconstruct and semantically understand full 360° scenes in open-world environments, without being limited by fixed vocabularies or narrow field-of-view inputs.

🎥 Teaser

📰 News

• 03/12/2026: Repository initialized. Please watch this repo for updates. Thank you for your interest in our work!

🙏 Acknowledgement

This project builds upon several outstanding open-source projects. We gratefully acknowledge the following key contributions.

• MonoScene, SGN, OVO, OneOcc, Spatial-Mamba.

🤝 Publication:

Please consider referencing this paper if you use the code or data from our work. Thanks a lot :)

@article{duan2026o3n,
  title={O3N: Omnidirectional Open-Vocabulary Occupancy Prediction},
  author={Mengfei Duan and Hao Shi and Fei Teng and Guoqiang Zhao and Yuheng Zhang and Zhiyong Li and Kailun Yang},
  journal={arXiv preprint arXiv:2603.12144},
  year={2026}
}