Fast Soft Color Segmentation

• Paper Link
• Supplimentary Video

This repository was developed as a part of an internship at Preferred Networks in the summer of 2019.

Usage

Training

• Prepare a train dataset
• Prepare CSV files for image paths and corresponding palette values, like sample.csv and palette_7_sample.csv.
• Run train.py with arguments.

Inference

Inference.ipynb shows a sample decomposition with apple.jpg. Run it in order from top. If you want to use other images, please change:

# image name and palette color values
img_name = 'apple.jpg'; manual_color_0 = [253, 253, 254]; manual_color_1 = [203, 194, 170]; manual_color_2 = [83, 17, 22]; manual_color_3 = [205, 118, 4]; manual_color_4 = [220, 222, 11]; manual_color_5 = [155, 24, 10]; manual_color_6 = [171, 75, 67];

manual_color_X means user-specified RGB values. If necessary, K-means algorithm (bottom of the notebook) give you these values.

Notes

• This is developed on a Linux machine running Ubuntu 16.04
• Distributed pretrained model is for 7 layer decomposition.
• The copyright of apple.jpg belongs to Adelle Chudleigh.

@InProceedings{Akimoto_2020_CVPR,
author = {Akimoto, Naofumi and Zhu, Huachun and Jin, Yanghua and Aoki, Yoshimitsu},
title = {Fast Soft Color Segmentation},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}