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.
• 2. 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.
• 3. 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).
• 4. 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.

1. Data downloading & preprocessing & color extraction algoritm
2. Data analysis and conclusions

Results