/DFSS-IQA

Code for "Deep Feature Statistics Mapping for Generalized Screen Content Image Quality Assessment".

Primary LanguagePythonMIT LicenseMIT

DFSS-IQA

Code for "Deep Feature Statistics Mapping for Generalized Screen Content Image Quality Assessment".
image

Environment

  • python=3.8.5
  • pytorch=1.7.1 cuda=11.0.221 cudnn=8.0.5_0

Running

  • Data Prepare
  • Download the SCID and SIQAD datasets into the path: ./DFSS-IQA/datasets/
  • We provide the pre-trained checkpoints here. You can download it and put the included files into the path: "./DFSS-IQA/DFSS_Release/models/".

  • Train:

    • For Intra-dataset:

    • SIQAD: python iqaScrach.py --list-dir='../sci_scripts/siqad-scripts-6-2-2/' --pro=split_id --resume='../models/siqad/checkpoint_latest.pkl' --dataset='IQA'

    • SCID: python python iqaScrach.py --list-dir='../sci_scripts/scid-scripts-6-2-2/' --pro=split_id --resume='../models/scid/checkpoint_latest.pkl' --dataset='SCID' --n-dtype=46

      • split_id: 0-9

    • For Cross-dataset:

    • SIQAD: python iqaScrach.py --list-dir='../sci_scripts/siqad-scripts-all/' --pro=0 --resume='../models/siqad-all/checkpoint_latest.pkl' --dataset='IQA'

    • SCID: python iqaScrach.py --list-dir='../sci_scripts/scid-scripts-all/' --pro=0 --resume='../models/scid-all/checkpoint_latest.pkl' --dataset='SCID' --n-dtype=46

  • Test:

    • Intra-dataset: python iqaIntraTest.py
    • Cross-dataset: python iqaCrossTest.py
    • Demo: python demo.py