Awesome

<p align="center">No Reference Opinion Unaware Quality Assessment of Authentically Distorted Images</p>

Environment

The code has been tested with:

• Python 3.6.13
• Pytorch 1.10.1
• Torchvision 0.11.2
• CUDA 11.3.1

Set up an environment with Python 3.6.13 and install all required libraries with the requirements.txt file. If you use conda, follow the instructions below:

Setting up conda environment

Execute the following lines one by one on the terminal.

conda create --name iqa_content_sep python=3.6.13
conda activate iqa_content_sep
conda install pytorch==1.10.1 torchvision==0.11.2 cudatoolkit=11.3 -c pytorch -c conda-forge
conda install --file requirements_extra.txt -c conda-forge

Training

Synthetic data pretraining

To train the model on the KADIS dataset, run the following:

python ./synthetic_pretraining.py --log

Authentic data fine-tuning

After synthetic data pre-training, please run the script below to fine-tune it further on 10k samples of the AVA dataset. The AVA dataset samples used for training are specified in the ./datasets/AVA_10k_set.csv file.

python ./authentic_finetuning.py --log --mi_contrastive

Performance evaluation

Pre-trained weights

Google Drive link for pre-trained weights:

• Model pre-trained on synthetic distortions
• Model finetuned on authentic distortions

Copy the files to the ./pre_trained_models/ folder.

Setting up pristine patches

Google Drive link for pre-selected pristine patches link. Copy the downloaded file to the dataset_images folder as ./dataset_images/pristine_patches_096_0.75_0.80.hdf5 .

Testing code

Sample testing code for evaluating the final model on different authentically distorted datasets.

python ./evaluate_model.py --dataset LIVEC --model_weights ./pre_trained_models/auth_ft_cd.pth --eval_result_dir ./results/auth_ft_cd/
python ./evaluate_model.py --dataset KONIQ --model_weights ./pre_trained_models/auth_ft_cd.pth --eval_result_dir ./results/auth_ft_cd/
python ./evaluate_model.py --dataset LIVEFB --model_weights ./pre_trained_models/auth_ft_cd.pth --eval_result_dir ./results/auth_ft_cd/
python ./evaluate_model.py --dataset CID --model_weights ./pre_trained_models/auth_ft_cd.pth --eval_result_dir ./results/auth_ft_cd/

Citation

If you find this work useful for your research, please cite our paper:

@InProceedings{iqa_content_sep,
    author    = {Babu, Nithin C. and Kannan, Vignesh and Soundararajan, Rajiv},
    title     = {No Reference Opinion Unaware Quality Assessment of Authentically Distorted Images},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2023},
    pages     = {2459-2468}
}

License

MIT License

Copyright (c) 2023 Nithin C Babu, Vignesh Kannan, Rajiv Soundararajan

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.