/deep-erqa

Video restoration assessment

Primary LanguageJupyter Notebook

Deep ERQA

NN-based Full-Reference metric for assessing image restoration quality.

Usage example

  1. Install requirements pip install -r requirements.txt
  2. Download weights: https://drive.google.com/file/d/11oo-EoJN5yVQ30m0PwBA4qgqpY3NBg3m/view?usp=sharing
  3. Unzip downloaded file
  4. Modify edge_similarity/test_benchmark.py for your needs and execute.
python edge_similarity/test_benchmark.py --gpu 0 ${MODEL_PATH}/best_model/model ${DATA_PATH} ${RES_PATH}

Train example

  1. Install requirements pip install -r requirements.txt
  2. Download fannet.zip from this link and unzip to ${DATA_PATH}
  3. Run the following command
python edge_similarity/train.py --gpu 0 ${DATA_PATH}