tobiascz / TeCNO

Table of Contents

1. About
2. Getting started
   1. Prerequisites
   2. Usage
3. Reference
4. Contact

About

TeCNO performs hierarchical prediction refinement with causal, dilated convolutions for surgical phase recognition and outperforms various state-of-the-art LSTM approaches!

Link to paper: TeCNO Paper

Getting started

Follow these steps to get the code running on your local machine!

Prerequisites

pip install -r requirements.txt

Usage

We are using the publicly available Cholec80 dataset. For training we split the videos into individual frames.

Stage 1 - Train Feature Extractor

Run:

python train.py -c modules/cnn/config/config_feature_extract.yml

This will train your feature extractor and in the Test Step it will extract for each Video the features of all images and save it as .pkl

Stage 2 - Train Temporal Convolutional Network

python train.py -c modules/mstcn/config/config_tcn.yml

Reference

@inproceedings{czempiel2020,
 author    = {Tobias Czempiel and
               Magdalini Paschali and
               Matthias Keicher and
               Walter Simson and
               Hubertus Feussner and
               Seong Tae Kim and
               Nassir Navab},
 title     = {TeCNO: Surgical Phase Recognition with Multi-Stage Temporal Convolutional
               Networks},
  booktitle = {Medical Image Computing and Computer Assisted Intervention - {MICCAI}
               2020 - 23nd International Conference, Shenzhen, China, October 4-8,
               2020, Proceedings, Part {III}},
  series    = {Lecture Notes in Computer Science},
  volume    = {12263},
  pages     = {343--352},
  publisher = {Springer},
  year      = {2020},
}

Contact

For any problems and question please open an Issue

Follow me on