Representation learning using rank loss for robust neurosurgical skills evaluation.
To install this repository and its dependent packages, run the following.
git clone https://github.com/myselfbritty/NeuroEval.git
cd NeuroEval
conda create --name NeuroEval # (optional, for making a conda environment)
conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.2 -c pytorch
pip install -r requirements.txt
For JIGSAWS dataset with the extracted visual features, please refer to TUSSA and their provided files here.
For NETS data, visit Project page.
The current paths assume these datasets to be in ./datasets/JIGSAWS
and ./datasets/NETS
respectively. You may follow the same directory structure or change the paths in the configs.
To train and test a model, make necessary changes in a config file and run the following command.
python3 main.py --config <config path> --log_dir <path to save logs>
For example, to run the default setting 4-fold validation on JIGSAWS, run the following.
python3 main.py --config configs/JIGSAWS_rank2/Knot_Tying/JIGSAWS_TVPE_4FOLD_Knot_Tying_0.json --log_dir logs/JIGSAWS_rank2
Similarly, to run the default setting train-test split on NETS, run the following.
python3 main.py --config configs/NETS_rank2/NETS_train_test.json --log_dir logs/NETS_rank2
This repository benefits greatly from TUSSA
and we thank its authors for this awsome work.
@inproceedings{baby2022representation,
title={Representation Learning Using Rank Loss for Robust Neurosurgical Skills Evaluation},
author={Baby, Britty and Chasmai, Mustafa and Banerjee, Tamajit and Suri, Ashish and Banerjee, Subhashis, and Arora, Chetan},
booktitle={2022 29th IEEE International conference on image processing (ICIP)},
year={2022},
organization={IEEE}
}