lxc0205 / curd-biqa

CURD-BIQA

Dependencies

conda create -n curd python=3.8.18
conda activate curdbiqa
pip install -r requirements.txt

Usage

Multiscale framework for BIQA

python framework.py <--curd> --method <method> --dataset <dataset> --index <indexs> --beta <betas>

Some available options:

• --dataset: Testing dataset, support datasets: koniq-10k | live | csiq | tid2013.
• --method: basic method.
• --curd: The flag of using curd (Using multiscale framework), switch off represents the original method.
• --index: The curd selected indexs, only predicting process used.
• --beta: The curd selected betas, only predicting process used.

Outputs:

• orignal mehthod and predicting processing: print PLCC and SRCC
• Usage of curd, get the layer scores: .\outputs\<method>\ multiscale outputs\<dataset>.txt

CURD-BIQA

python curd_iqa.py --method <method> --inputFileSet <input file names> --outputFile <output file name> --save_num <save_num>

Some available options:

• --method: basic method.
• --inputFileSet: input file names in multiscale outputs folder, support names: koniq-10k.txt | live.txt | csiq.txt | tid2013.txt.
• --outputFile: output file names.
• --save_num: Save numbers, default number is 5000000.

Outputs:

• output file: .\outputs\<method>\curd outputs\<outputFile>.txt