A library to train, test and develop Video Super Resolution architectures. The framework is based on Hydra and offers a variety of SoA architectures (Real Basic VSR, VRT) already implemented and ready to be trained.
conda create -n vsrlab python=3.11
conda activate vsrlab
pip3 install -r requirements.txt
git clone https://github.com/santurini/vsrlab.git
cd vsrlab && pip3 install .
python train.py +experiment=basic
# to resume a training -> change 'restore' and 'restore_opt' fields in experiment config file
python test.py +experiment=test cfg_dir=path/to/checkpoints_dir/experiment_name
# checkpoints directory should contain 'checkpoint/last.ckpt' and 'config.yaml'