README.md

General Data Process

• We only use random crop for data augmentation.
• Remember reset the normalization based on your need.
• We use opencv (cv2) to read and process images.

How To Prepare Training Data

1. BAPPS 2afc dataset. Download using BAPPS offical script.

Modify configurations in codes/options/train_test_yml/train_our_IQA.yml when training, e.g., train_root, valid_root and train_valid (Choose which part of BAPPS 2afc dataset become training data).

2. PIPAL dataset. Download Training, Validation and Testing (2021-03-01).

For PIPAL training dataset, you could obtain four files named Distortion_1, Distortion_2, Distortion_3 and Distortion_4. Please create one file which only contains these four subfiles.
Modify configurations in codes/options/train_test_yml/train_our_IQA.yml when training, e.g., mos_root, ref_root and dis_root.

For PIPAL validation (Ref_valid / Distortion_valid), place them in the one file.
When you want to test your model. Modify configurations in codes/options/train_test_yml/test_IQA.yml when testing, e.g., ref_root and dis_root.

3. The ideal file structure is shown below.

Important! The "Distortion" file should only contains four subfiles described here. Otherwise, you should modify function image_combinations in codes/data/data_util.py.

PIPAL
   │   
   └───validation (Public)
   │   |
   │   └───Reference_valid
   │   │     |A0000.bmp
   │   │     |A0003.bmp
   │   │     |...
   │   │            
   │   └───Distortion_valid
   │         |A0000_10_00.bmp
   │         |A0000_10_01.bmp
   │         |...
   │   
   └───training (Public)
       │
       └───MOS_Scores_train
       │     |A0001.txt
       │     |A0002.txt
       │     |...
       │ 
       └───Reference_train
       │     |A0000.bmp
       │     |A0003.bmp
       │     |...
       │            
       └───Distortion 
             └───Distortion_1
             │       | ...
             └───Distortion_2
             │       | ...
             └───Distortion_3
             │       | ...
             └───Distortion_4
                     | ...