Skip to content


Repository files navigation


Tool for segmenting endometrial implants in laparoscopic surgery videos.


Demonstrating partial results of ongoing thorough studies on endometriosis segmentation, this web-based demo provides a user with the tools for segmenting endometrial implants in laparoscopic surgery videos. Following figure provides an overview of the tool's processing approach:


  1. First a user provides the tool with an input video
  2. The video is analyzed using one of several provided segmentation models
  3. The analyses are utilized to additionally add a summarization bar that indicates frame-by-frame averaged segmentation confidences, while also marking the current video position.


We kindly ask you to cite following papers if you include our work in your research:

Post-surgical Endometriosis Segmentation in Laparoscopic Videos

  author    = {Andreas Leibetseder and
               Klaus Schoeffmann and
               J{\"{o}}rg Keckstein and
               Simon Keckstein},
  title     = {Endometriosis detection and localization in laparoscopic gynecology},
  journal   = {Multim. Tools Appl.},
  volume    = {81},
  number    = {5},
  pages     = {6191--6215},
  year      = {2022},
  url       = {},
  doi       = {10.1007/s11042-021-11730-1},
  timestamp = {Thu, 03 Mar 2022 09:23:23 +0100},
  biburl    = {},
  bibsource = {dblp computer science bibliography,}

  author    = {Andreas Leibetseder and
               Klaus Schoeffmann and
               J{\"{o}}rg Keckstein and
               Simon Keckstein},
  title     = {Post-surgical Endometriosis Segmentation in Laparoscopic Videos},
  booktitle = {18th International Conference on Content-Based Multimedia Indexing,
               {CBMI} 2021, Lille, France, June 28-30, 2021},
  pages     = {1--4},
  publisher = {{IEEE}},
  year      = {2021},
  url       = {},
  doi       = {10.1109/CBMI50038.2021.9461900},
  timestamp = {Tue, 29 Jun 2021 17:37:02 +0200},
  biburl    = {},
  bibsource = {dblp computer science bibliography,}



  • Python 3.8.2

    Install on system or optionally create and activate virtual environment:

    • via Anaconda

      conda create --name EndometriosisSegmentationTool python=3.8.2
      conda activate EndometriosisSegmentationTool
    • via Python

      python -m venv /path/to/virtual/envs/EndometriosisSegmentationTool
      source /path/to/virtual/envs/EndometriosisSegmentationTool/bin/activate
  • CUDA 10.2 for capable Nvidia GPUs (optional) Follow installation instructions for your particular OS.

  • PyTorch>=1.6.0 and TorchVision>=0.7.0 (optionally for CUDA 10.1)

    • using conda:
      conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.2 -c pytorch
    • using pip:
      python -m pip install torch==1.6.0 torchvision==0.7.0
  • Detectron2 v0.4 use pre-built files (Linux only) or extract to any folder and build:

    python -m pip install -e detectron2-0.4

    Hint: Although Windows 10 is not officially supported, building should work anyways. If problems occur, please refer to conansherry/detectron2#2.

  • Required Python packages

    python -m pip install -r requirements.txt


Pre-trained models can be downloaded from the ENdometrial Implants Dataset (ENID) Homepage or Zenodo page.


python -h
usage: [-h] -i IN -m MODEL [-o OUT]

optional arguments:
  -h, --help            show this help message and exit
  -i IN, --in IN        path to video or input folder containing videos
  -m MODEL, --model MODEL
                        path to input model or root folder containing multiple model subfolders with their
                        respectie config.yaml files
  -o OUT, --out OUT     path to output folder (default: [IN_PATH]_out)

Django Application

For ease of use a Python Django Application provides graphical UI for analyzing single videos.


Make sure is set up to include the correct path to Python 3.x (aliases do not work in bash files).


Run Server


Optional: to run tool using a specific GPU device (e.g. 1) set following environment variable:


Clear App

WARNING: this will clear the database and remove all previously conducted analyses.



No description, website, or topics provided.







No releases published