Skin-color extraction and classification

This programm allow you to extract skin color from face and classify it to the arbitraty number of clusters.

It was done as a part of my studies in London in 2019

The pipeline of my method may be divided into four primary segments:

1. Face detection - obtaining bounding box around the face.
1. Apply k-Means Clustering to the face on the HSV color space to obtain the most dominant color which is assumed to be the skin color.
1. The center of the prevalent cluster would represent the color of a person. So, I collect those colors for a set of images (~ 20k+ using UTKFace dataset).
1. After this I apply k-Means clustering algorithm to the set of collected colors and look for an optimal number of clusters that would give a meaningful result.

Project structure

Project consist of 2 jupyter notebooks.

Data downloading & preprocessing & color extraction algoritm
Data analysis and conclusions

Results