A PyTorch implementation of the paper Symmetric Dilated Convolution for Surgical Gesture Recognition.

This implementation uses Python 3.6, Pytorch 1.2.0, cudatoolkit 9.2. We recommend to use conda to deploy the environment

Install with conda:

conda env create -f environment.yml
conda activate sd-conv

This code is tested on the suturing task of JIGSAWS dataset. We use the same data features as Colin Lea. Please request the Spatial-CNN feature from Colin and put the features (39 extracted video features) intosd-conv/JIGSAWS/data folder.

We also test our method on 50Salads and GTEA datasets, but do not formally benchmark or claim any results.

To train, test, and visulize the result for one split run:

cd sd-conv
python train.py

The visualization result is located in sd-conv/visualization

If you find our work helpful, please consider citing:

@inproceedings{zhang2020symmetric,
    title={Symmetric Dilated Convolution for Surgical Gesture Recognition},
    author={Zhang, Jinglu and Nie, Yinyu and Lyu, Yao and Li, Hailin and Chang, Jian and Yang, Xiaosong and Zhang, Jian Jun},
    booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
    pages={409--418},
    year={2020},
    organization={Springer}
}