In addition to the .py files that contain the relevant code itself, we also provide two Jupyter-Notebooks to show-case the code.
The code relies on three main packages for dataloading:
- albumentations (augmentations and other transforms)
- pytorch (general dataset structure)
- pycocotools (mscoco-related stuff)
Furthermore, for the compositing of per-fabric datasets into bigger datasets, it uses:
- pytorch-lightning
In order to get started, please set-up an appropriate python environment and install all dependencies. To do this automatically, you can run
conda env create -f environment.yml
Start-up a local Jupyter server via
conda activate olp
jupyter lab