Attention-Guided Neural Networks for Full-Reference and No-Reference Audio-Visual Quality Assessment

Description

ANNAVQA code for the following papers:

Y. Cao, X. Min, W. Sun and G. Zhai, "Attention-Guided Neural Networks for Full-Reference and No-Reference Audio-Visual Quality Assessment," in IEEE Transactions on Image Processing, vol. 32, pp. 1882-1896, 2023, doi: 10.1109/TIP.2023.3251695.
Y. Cao, X. Min, W. Sun and G. Zhai, "Deep Neural Networks For Full-Reference And No-Reference Audio-Visual Quality Assessment," 2021 IEEE International Conference on Image Processing (ICIP), 2021, pp. 1429-1433, doi: 10.1109/ICIP42928.2021.9506408.

Download the LIVE-SJTU Database
Saliency Detection You should first run sal_position.m in Matlab to get SJTU_position.mat. You need to modify the databasePath into your save path of the LIVE-SJTU Database.
```
cd Saliency model
sal_model
```
Extract video features
```
cd train
python video_CNNFeatures.py
```
Extract audio features
```
cd train
python audio_CNNFeatures.py
```
Train the FR model
```
cd train
python ANNAVQA_ref.py
```
Train the NR model
```
cd train
python ANNAVQA_noref.py
```