This repository contains the official implementation for the paper:

Self-Supervised Dual Contouring
Ramana Sundararaman, Roman Klokov, Maks Ovsjanikov
arXiv:2405.18131 [cs.CV]

The paper can be found on arXiv.

To install the required dependencies, run the following command:

pip install -r requirements.txt

To train a new model, you can run the train.py script with a configuration file. For example:

python train.py --config configs/config_file.yaml --exp_name my_experiment

To evaluate a trained model on a dataset, use the single_eval.py script. You will need to provide paths to the decoder weights and the data directory.

python single_eval.py --dec_weight /path/to/your/model.th --data_dir /path/to/your/data --save_dir ./results

Check the script for more options.

An example configuration file can be found at Configs/example.yaml. This file contains all the parameters for training and data loading.

If you find our work useful, please consider citing our paper:

@misc{sundararaman2024selfsupervised,
      title={Self-Supervised Dual Contouring}, 
      author={Ramana Sundararaman and Roman Klokov and Maks Ovsjanikov},
      year={2024},
      eprint={2405.18131},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}