NEMG-IQA

Prerequisites

Ubuntu 18.04

Python 3.6.10

The sufficient requirements of libs in Python could be seen in ./env.list.txt

Generate Json score files.

You can generate score files by running prepare_score_file.py in ./DatasetScores
```
python prepare_score_file.py
```
The default setting would generate 10 splits per dataset(CSIQ,TID2013,LIVE).
Configure your setting

You should configure your score file and dataset path before training. Please open the ./configuration/{dataset}_config.json and change the image_dir and score_file term

Training

You could train the model by the following command:

python trainer.py --config configuration/{dataset}_config.json

The Resnet-50 is adapted from Pytorch model zoo