Crowd-Powered Photo Enhancement Featuring an Active Learning Based Local Filter

This is the official implementation of the paper in TCSVT2023.

Requirements

• Python 3.5.9
• MATLAB R2016b

To use MATLAB functions from Python, at the MATLAB command prompt,

cd (fullfile(matlabroot,'extern','engines','python'))
system('python setup.py install')

To install the Python libraries,

pip install --upgrade pip
pip install -r requirements.txt
pip install ./sequential-line-search_my
pip install megengine -f https://megengine.org.cn/whl/mge.html

Experiments using BIQME

1. Different number of pixels L --num_pixels.

python evaluation/RunExperiment.py --num_pixels 4 --mode BIQME

2. Without active learning --woactivelearning or the illumination map --woilluminationmap.

python evaluation/RunExperiment.py --mode BIQME --woactivelearning
python evaluation/RunExperiment.py --mode BIQME --woilluminationmap

3. Previous local filters. --filter_type can be set to global, graduated, elliptical, cubic10, and cubic20.

python evaluation/RunExperimentLPF.py --mode BIQME --filter_type graduated

The BIQME scores are saved as .npy files, and the file name is printed at the end of the process. By setting the file names in BIQME_scores of draw_graph.py and executing draw_graph.py, the graph of the BIQME scores is output.

Experiments on Amazon Mechanical Turk

To conduct experiments on Amazon Mechanical Turk (AMT), config.py needs to be set first. For AMT's API, please create an Amazon Web Service account, get a pair of an access key ID and a secret access key, and set them in "aws_access_key_id" and "aws_secret_access_key" of AMTAPI_config.

To publish images to crowd workers, a server with HTTP access is needed. Please set the server's IP address, port number, user name, and URL in "sshIP", "sshPort", "sshUsername", and "httpURL" of fileServer_config. "sshDirectory" should be set to the directory pointed by "httpURL".

By setting AMT_config["sandbox"] as True, you can check the interface for the crowd workers in the sandbox environment without paying the fee.

1. Our method.

python evaluation/RunExperiment.py --mode AMT --image_names 0,1,2,3,4

2. Sequential Line Search.

python evaluation/RunExperimentSLS.py --mode AMT --image_names 0,1,2,3,4

Demo

You can adjust a sequence of single sliders by yourself.

python evaluation/RunExperiment.py --mode self --image_names 19

Results

Results of our method and compared previous methods can be downloaded from here.

Citation

If you find our research useful in your research, please consider citing:

@article{kosugi2023crowd,
    title={Crowd-Powered Photo Enhancement Featuring an Active Learning Based Local Filter},
    author={Kosugi, Satoshi and Yamasaki, Toshihiko},
    journal={IEEE Transactions on Circuits and Systems for Video Technology},
    volume={33},
    number={7},
    pages={6493--6501},
    year={2023}
}