myovision

Description

This is the sub-repository of the main project myovision. It's main purpose is to provide various utilities for the data collection process of the myovision project.

Installation

git clone git@github.com:Noza23/myovision-data-utils.git
cd myovision-data-utils
pip install -r requirements.txt

Repository contains following utilities:

1. clean_masks.py The module is used to remove redundant masks manually from the given image. It additionally provides a visualization of the masks drawn on the image to help the user decide which masks to keep. See: clean_masks.sh for more details about parameters and usage of the script. Simple execution with:

   sh clean_masks.sh

2. merge_masks.py The module is used to merge masks from different sources into a single mask. In our case between:

   • Masks as RLE encoded json files generated by the annotation tool
   • .roi files generated by free-hand annotation in ImageJ.

   Check python3 merge_masks.py --help for more details about parameters and usage. Simple execution with:

   python3 merge_masks.py --masks /path/to/json_masks \
                            --roi /path/to/roi_masks \
                            --output /path/to/output

3. data_generator.py The module takes directory where images and corresponding masks are located:

   • Cuts each image:mask pair into patches of the given size.
   • Performs various data augmentation configured in data_generator.yaml.
   • Saves resulting data prepared for training in the given output directory.

   - Note: In the Directory you should have following naming convention:
        images: *.png
        masks: *_mask.json

   Fill in the data_generator.yaml with the desired parameters and execute the script with:

   python3 data_generator.py