This repo contains a basic data loading class DataLoader
designed to read images and masks from a directory structure and hand them off to a neural network in batches. It also contains some example data as well as visualization tools and some basic unit tests. See the Jupyter notebook for a run-through of some visualizations and working example of feeding the data to a Keras UNet.
jliemansifry/spine_data_pipeline
Pipeline demo for feeding data to a neural network for semantic segmentation
Jupyter Notebook