On the sufficient condition for solving the Gap-filling problem using Deep Convolutional Neural Networks
This repository contains the code to reproduce the experiments.
All necessary requirements are stated in environment.yml and may be installed using conda.
For example, to create a new environment and install the requirements:
$ conda env create --name dcnn-gap-filling --file environment.ymlThe tessellation script can be found in tessellation-generator/tessellator.py.
The parameters for the datasets can be found at tessellations_test/params.json and tessellations_train/params.json.
The datasets can be (re-)generated by running the script ./generate_datasets.sh.
An additional script to view some general statistics about the generated datasets can be found in analyze_tessellations.
The muscle fibre dataset can be downloaded at https://github.com/AnikaNovikov/SLCV.
Code which is reused by the experiments is in the folder unet.
The implementation is adapted from https://github.com/jvanvugt/pytorch-unet.