JunFu1995 / AHGCN

Adaptive Hypergraph Convolutional Network for No-Reference 360-degree Image Quality Assessment, ACM MM 2022

Adaptive Hypergraph Convolutional Network for No-Reference 360-degree Image Quality Assessment

TODO

• Code release on the CVIQD dataset
• Release the code for the OIQA dataset

Introduction

For the implementation details of AHGCN, please see the file "./model/cviqd.py".

Train and Test

First, download datasets from Here.

Second, modify the datapath in line 53 in the datasets/cviqd.py:

Thrid, run the following command to train and test the model

python train_cviqd_shell.py

The best model and its corresponding predictions are saved in the directory "./save/" and the directory "./mat/", respectively.

We have uploaded the checkpoint of the best model retrained on a single NVIDIA RTX 4090 GPU in the directory "./save/best_cviqd"

Citation

Please cite the following paper if you use this repository in your reseach.

@inproceedings{fu2022adaptive,
  title={Adaptive hypergraph convolutional network for no-reference 360-degree image quality assessment},
  author={Fu, Jun and Hou, Chen and Zhou, Wei and Xu, Jiahua and Chen, Zhibo},
  booktitle={Proceedings of the 30th ACM International Conference on Multimedia},
  pages={961--969},
  year={2022}
}

Contact

For any questions, feel free to contact: [e-mail]