MC360IQA

MC360IQA: A Multi-channel CNN for Blind 360-degree Image Quality Assessment

Usage

If you want to train the code on your database (e.g. CVIQ database ):

First, prepare the database

cd equi2cubic
ConvertCVIQtoCubic.m

Then

CUDA_VISIBLE_DEVICES=0 python train.py \
--num_epochs 10 \
--batch_size 40 \
--database CVIQ \
--data_dir /DATA/CVIQcubic \
--filename_train CVIQ/CVIQ_train.csv \
--filename_test CVIQ/CVIQ_test.csv \
--snapshot  /DATA/ModelFolder/VRIQA \
--cross_validation_index 1

If you want to test the trained model on the test set:

CUDA_VISIBLE_DEVICES=1 python test.py \
--database CVIQ \
--data_dir /DATA/CVIQcubic \
--filename_test CVIQ/CVIQ_test.csv \
--snapshot  /DATA/ModelFolder/VRIQA/CVIQ/1/CVIQ.pkl

If you just want to evaluate the quality of an equirectangular image:

CUDA_VISIBLE_DEVICES=0 python test_on_equirectangular.py \
--filename images/1.png \
--snapshot  /DATA/ModelFolder/VRIQA/CVIQ/1/CVIQ.pkl

Model

You can download the trained model via:

CVIQ: google drive baidu yun 提取码：5muh

OIQA: google drive baidu yun 提取码：39we

We recommend you to use the model trained on the OIQA database since it is more robust.

Citation

If you find this code is useful for your research, please cite:

@article{sun2019mc360iqa,
  title={MC360IQA: A Multi-channel CNN for Blind 360-degree Image Quality Assessment},
  author={Sun, Wei and Min, Xiongkuo and Zhai, Guangtao and Gu, Ke and Duan, Huiyu and Ma, Siwei},
  journal={IEEE Journal of Selected Topics in Signal Processing},
  volume={14},
  number={1},
  pages={64-77},
  year={2020},
  publisher={IEEE}
}

Acknowledgement

1. https://github.com/rayryeng/equi2cubic
2. https://github.com/pepepor123/equirectangular-to-cubemap
3. https://github.com/vztu/RAPIQUE