An example project using PipelineX, Kedro, OpenCV, Scikit-image, and TensorFlow/Keras for image processing.

Pipeline visualized by Kedro-viz

• conf
  • YAML config files for PipelineX project
• data
  • empty folders (output files will be saved here)
• logs
  • empty folders (log files will be saved here)
• src
  • empty_area.py
    • The algorithm to estimate empty area ratio
  • roi.py
    • Supplementary algorithm to compute ROI (Region of Interest) from segmentation image
  • semantic_segmentation.py
    • Semantic segmentation using PSPNet model pretrained with ADE20K dataset

$ pip install pipelinex opencv-python scikit-image ocrd-fork-pylsd Pillow pandas numpy requests kedro mlflow kedro-viz

Note: mlflow and kedro-viz are optional.

$ pip install "tensorflow<2" keras-segmentation Keras 

$ pip install "tensorflow>=2.0.0" Keras 
$ pip install git+https://github.com/Minyus/image-segmentation-keras.git

$ git clone https://github.com/Minyus/pipelinex_image_processing.git
$ cd pipelinex_image_processing

$ python main.py

As configured in catalog.yml, the following 2 images will be downloaded by http requests and then processed using opencv-python, scikit-image, and ocrd-fork-pylsd packages.

$ mlflow server --host 0.0.0.0 --backend-store-uri sqlite:///mlruns/sqlite.db --default-artifact-root ./mlruns/experiment_001

Experiment logs in MLflow's UI

• Python 3.6.8

This project was created from the GitHub template repository at https://github.com/Minyus/pipelinex_template

To use for a new project, fork the template repository and hit Use this template button next to Clone or download.