UniForge3D: Joint 3D Reconstruction and Generation in a Shared Anchor-Guided Space
Wenjue Chen1,*, Wenyue Chen1,2,*, Heliang Zheng2, Rongfei Jia2, Yuan Liu3,†, Meng Li1,†
1 Peking University 2 Math Magic 3 The Hong Kong University of Science and Technology
* Equal contribution. † Corresponding authors.
UniForge3D generates complete 3D assets from sparse, unposed multi-view images. It jointly models reconstruction and generation in a shared anchor-guided space, then uses structure-guided view selection to preserve observed details.
Installation
git clone https://github.com/xibushu/UniForge3D.git
cd UniForge3D
bash setup.sh
conda activate uniforge3dInference
Place the views of each object directly in a separate folder under assets/:
assets/
├── case01/
│ ├── 001.png
│ ├── 002.png
│ ├── 003.png
│ └── 004.png
└── case02/
└── ...
Run all cases using the pretrained model:
python infer.py
Acknowledgements
Our code is based on these wonderful works:
Citation
If you find UniForge3D useful in your research, please cite our paper:
@article{chen2026uniforge3d, title = {UniForge3D: Joint 3D Reconstruction and Generation in a Shared Anchor-Guided Space}, author = {Chen, Wenjue and Chen, Wenyue and Zheng, Heliang and Jia, Rongfei and Liu, Yuan and Li, Meng}, journal = {arXiv preprint arXiv:2610.15357}, year = {2026} }