This paper has been accepted by ICCV 2023

By Ziyue Feng, Liang Yang, Pengsheng Guo, and Bing Li.

Project Page: cvrecon.ziyue.cool

Video

Dear readers:

Apologize for late release of the code, I have been too busy recently so still have not got time to clean up the code.

I hope this initial release could give you some idea about how the CVRecon works. The implementation is based on the Cost Volume of "SimpleRecon" and the framework of "VoRTX".

I will clean up the code as soon as I got time.

Dependencies

conda create -n cvrecon python=3.9 -y
conda activate cvrecon

conda install pytorch torchvision cudatoolkit=11.3 -c pytorch

pip install \
  pytorch-lightning==1.5 \
  scikit-image==0.18 \
  numba \
  pillow \
  wandb \
  tqdm \
  open3d \
  pyrender \
  ray \
  trimesh \
  pyyaml \
  matplotlib \
  black \
  pycuda \
  opencv-python \
  imageio

sudo apt install libsparsehash-dev
pip install torchsparse-v1.4.0 

pip install -e .

Data

The ScanNet data should be downloaded and extracted by the script provided by the authors.

To format ScanNet for cvrecon:

python tools/preprocess_scannet.py --src path/to/scannet_src --dst path/to/new/scannet_dst

In config.yml, set scannet_dir to the value of --dst.

To generate ground truth tsdf:

python tools/generate_gt.py --data_path path/to/scannet_src --save_name TSDF_OUTPUT_DIR
# For the test split
python tools/generate_gt.py --test --data_path path/to/scannet_src --save_name TSDF_OUTPUT_DIR

In config.yml, set tsdf_dir to the value of TSDF_OUTPUT_DIR.

Training

python scripts/train.py --config config.yml

Parameters can be adjusted in config.yml. Set attn_heads=0 to use direct averaging instead of transformers.

Inference

python scripts/inference.py \
  --ckpt path/to/checkpoint.ckpt \
  --split [train / val / test] \
  --outputdir path/to/desired_output_directory \
  --n-imgs 60 \
  --config config.yml \
  --cropsize 96

Evaluation

Refer to the evaluation protocal by Atlas and TransformerFusion

Citation

@misc{feng2023cvrecon,
  title={CVRecon: Rethinking 3D Geometric Feature Learning For Neural Reconstruction}, 
  author={Ziyue Feng and Leon Yang and Pengsheng Guo and Bing Li},
  year={2023},
  eprint={2304.14633},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}