TensoFlow: Tensorial Flow-based Sampler for Inverse Rendering

[Paper]

TensoFlow: Tensorial Flow-based Sampler for Inverse Rendering,
Chun Gu, Xiaofei Wei, Li Zhang, Xiatian Zhu
Shanghai Innovation Institute; School of Data Science, Fudan University; University of Surrey
CVPR 2025

Official implementation of "TensoFlow: Tensorial Flow-based Sampler for Inverse Rendering".

🛠️ Pipeline

⚙️ Installation

git clone https://github.com/fudan-zvg/tensoflow.git

conda create -n tensoflow python=3.10
conda activate tensoflow

pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu121
pip install -r requirements.txt
pip install git+https://github.com/ashawkey/raytracing.git

📦 Dataset

Please follow TensoSDF to download the TensoSDF dataset and the ORB dataset. Put them under the nerf_data folder. You may need to change the data path in the config files.

TensoSDF synthetic dataset

SCENE=compressor
DEVICE=0
# geometry reconstruction
bash run_occ_syn_shape.sh $SCENE $DEVICE
# material estimation
bash run_occ_syn_material.sh $SCENE $DEVICE

ORB dataset

SCENE=teapot
DEVICE=0
# geometry reconstruction
bash run_occ_orb_shape.sh $SCENE $DEVICE
# material estimation
bash run_occ_orb_mat.sh.sh $SCENE $DEVICE
# See an example of evaluating ORB dataset in eval_orb_relight.sh

🛠️ Visualization

📜 BibTeX

@inproceedings{gu2025tensoflow,
  title={TensoFlow: Tensorial Flow-based Sampler for Inverse Rendering},
  author={Gu, Chun and Wei, Xiaofei and Zhang, Li and Zhu, Xiatian},
  booktitle={CVPR},
  year={2025},
}