/misp

Medical Image Segmentation Pipeline

Primary LanguagePythonMIT LicenseMIT

misp

Medical Image Segmentation Pipeline

Note: DemoDataset folder in this repo is just to let you know how I organized my training data, therefore there is no "real" file inside.

Also, this folder structure is used as a premise in all my related scripts.


Requirements

  1. OpenSlide

    If you're using Windows, remember to put it's bin directory in your Environment Variables.

  2. openslide-python 1.1.1

    pip install openslide-python==1.1.1
  3. PyTorch 1.1.0

  4. Other requirements will automatically be installed when you install misp.


Installation

Install misp:

pip install misp

Q & A:

  1. How to create a subset of the original dataset for faster iterations?

Below is an example which I used to create my "smaller version" dataset.

python create_tiny_dataset.py -s CRCdataset/train -d tinyCRC/train -n 600

python create_tiny_dataset.py -s CRCdataset/valid -d tinyCRC/valid -n 100