This repository contains the implementation of the paper.

If you find our work useful, Please give us a star 🌟!

MagicDrive3D: Controllable 3D Generation for Any-View Rendering in Street Scenes
Ruiyuan Gao¹, Kai Chen², Zhihao Li³, Lanqing Hong^{3^}, Zhenguo Li³, Qiang Xu^{1^
1}CUHK ²HKUST ³Huawei Noah's Ark Lab
^{^}Corresponding Authors

The authors are all busy in other tasks and do not have time for code release. If you want to request the source code (DRAFT) for scene resonstruction, please send email to rygao.me@gmail.com. There will not be usage support. Thanks for your understanding.

As for upgrades in video generation, we will merge the implementation to MagicDrive-t (i.e., video branch of MagicDrive).

@inproceedings{gao2026magicdrive3d,
  title={{MagicDrive3D}: Controllable 3D Generation for Any-View Rendering in Street Scenes},
  author={Gao, Ruiyuan and Chen, Kai and Li, Zhihao and Hong, Lanqing and Li, Zhenguo and Xu, Qiang},
  booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  year={2026}
}
@inproceedings{gao2023magicdrive,
  title={{MagicDrive}: Street View Generation with Diverse 3D Geometry Control},
  author={Gao, Ruiyuan and Chen, Kai and Xie, Enze and Hong, Lanqing and Li, Zhenguo and Yeung, Dit-Yan and Xu, Qiang},
  booktitle = {International Conference on Learning Representations},
  year={2024}
}

We adopt the following open-sourced projects:

• MagicDrive: controllable street view video generation
• gaussian-splatting: 3D reconstruction
• ZoeDepth: monocular metric depth estimation