RoomDesigner: Encoding Anchor-latents for Style-consistent and Shape-compatible Indoor Scene Generation

3DV 2024 , arxiv preprint

Yiqun Zhao, Zibo Zhao, Jing Li, Sixun Dong and Shenghua Gao

Demo Results

Room-mask condtioned Scene generation

All the furniture are generated from our model (not retrievel from CAD library)

Bedroom results

Livingroom results

Dininingroom results

Scene Editing results

see the dinining chair

This editing was implemented by 1). edit single object with anchor-latents 2). Scene completion for scene editing. Please see the paper for more details.

Origin Scene	Edited Scene

Origin Scene	Edited Scene

Note

Model code is realesed! However, there are still some different configs and dataprocessing code to be released.

Feel free to contact: goatzyq@163.com

Acknowlegement

The code is mainly reference to:

• ATISS
• 3DILG

we thank the authors for their contributions.

Citation

If you find our project helpful in your research, please consider citing

@inproceedings{roomdesigner2024,
  title={RoomDesigner: Encoding Anchor-latents for Style-consistent and Shape-compatible Indoor Scene Generation},
  author={Yiqun Zhao and Zibo Zhao and Jing Li and Sixun Dong and Shenghua Gao},
  booktitle={Proceedings of the International Conference on 3D Vision (3DV)},
  year={2024}
}