# AVQA

**Repository Path**: lin_wei_hung/AVQA

## Basic Information

- **Project Name**: AVQA
- **Description**: Video-Audio Quality Assessment
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-12-27
- **Last Updated**: 2022-01-15

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# README

The paper is stored at https://gitee.com/lin_wei_hung/AVQA/blob/main/paper.pdf

## Run the Code

Train the model
```bash
# matlab -nosplash -nodesktop -r sal_position
# sal_position ./dis_test.yuv 1080 1920
# sal_position ./boxing/Boxing_QP50.yuv 1080 1920
python mytrain.py
```


Test model
```bash
python FR_LS_test.py --ref_video_path='./boxing/Boxing_QP16.yuv' --dis_video_path='./boxing/Boxing_QP50.yuv' --dis_audio_path='./boxing/Boxing_8kbps.wav' --ref_audio_path='./boxing/Boxing_128kbps.wav' --frame_rate=24
```


Use CAM to generate sampling positions
```bash
python detection.py

python cam.py --image-path /DATA3_DB7/data/weixionglin/AVQA/boxing/Boxing_QP16/Boxing_QP16-010.jpeg
python cam.py --image-path ../boxing/Boxing_QP16/Boxing_QP16-010.jpeg
python cam.py --image-path ../car/Car_QP16/Car_QP16-001.jpeg
```

Decode YUV to mp4, slice img from mp4
```bash
ffmpeg -s 1920x1080 -i Goose_QP50.yuv out3.mp4
ffmpeg -i Goose_QP35.mp4 -r 1 -q:v 2 -f image2 Goose_QP35-%03d.jpeg

ffmpeg -s 1920x1080 -i Boxing_QP16.yuv Boxing_QP16.mp4
ffmpeg -i Boxing_QP16.mp4 -r 1 -q:v 2 -f image2 Boxing_QP16-%03d.jpeg

ffmpeg -s 1920x1080 -i Car_QP16.yuv Car_QP16.mp4
ffmpeg -i Car_QP16.mp4 -r 1 -q:v 2 -f image2 Car_QP16-%03d.jpeg
```

## Requirments
Specify the required package, maybe i will remove them after the project is finished.

```bash
pip install sk-video

strings /usr/lib/x86_64-linux-gnu//libstdc++.so.6 | grep CXXABI


pip install --upgrade pandas
```

Dataset Location
LIVE-SJTU_AVQA.zip -> /DATA7_DB7/data/weixionglin
LIVE-SJTU_AVQA -> /DATA3_DB7/data/weixionglin