This project implements a deep learning architecture for unpaired image enhancement using reinforcement learning.
Run ./init <--i> <--gcp> with an optional -i argument to install the dependencies for the Python virtual environment and
an option --gcp flag to download the dataset from the GCP Storage bucket.
Run pytest from the root directory to run all unit tests.
To run the neural net model, use
python main.py [-h] [--checkpoint-dir] [--device] {train, test, evaluate, performance} ...
Use the -h or --help flags to view optional arguments for each command of {train, test, evaluate, performance}. For example python main.py train -h.
With learned model weights in the specified checkpoint-dir, run
python main.py evaluate --image-path IMAGE_PATH
where IMAGE_PATH is the path to any image stored locally to run the given image through the generator and display the resulting edits.
