Skip to content

My theisis which used ML to build a set of makeup recommendation based on users photo

Notifications You must be signed in to change notification settings

roryoreilly/thesis-make-up-recommender

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Make-Up Recommendation App

This repo contains the report and source code used to create my android make-up application. The application works by taking a photo of the user, analysing the skin tone, hair colour, eye colour, and the face shape using a machine learning model. With these features a set of makeup is recommended to the user that would most suit them.

Repo Structure

The repo is broken down in the report that was submitted for thesis completion and the source code which was used as the basis of the project

Snippet of the Thesis

4.1 Software Overview

software overview

As the app is turned on a database of the makeup products is created. There are 800 products, in total, in the app. Each of these are taken from the Mac range and contain different information regarding each product.

The app begins by getting an image of the user. This image is acquired as mentioned in the previous section. The image is passed to the next activity in the app.

The image is set to the correct format and sent to the Face++ servers. Face++ analyses the image and extract data on the face. This data includes the 83 landmark points on the face.

The result is returned from Face++ and the data, as well as the user image, is used to extract attributes for each of the features. The colour features, skin complexion, eyes, hair, and lips, will get a HSB value as it’s attribute. There will be six different attributes extracted for the face shape.

The features are classified using the attributes for each. They are classified using either Weka, in the case of eye colour and face shape, or a nearest neighbour algorithm in the case of skin complexion and hair colour. The user is able to reselect the correct classification in the cases where the app is wrong.

The occasion is selected by the user after the feature classification. There are three occasions, Night out, daytime, and natural. The occasion selection as well as the feature classes are passed to the next activity in the app, where the makeup is recommended.

Finally, the makeup is recommended and presented to the user. Depending on the occasion there are 5 to 15 products recommended for use. Each of the product recommendations are made individually so that they will be suited for the user and are complementary to each other.

About

My theisis which used ML to build a set of makeup recommendation based on users photo

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages