Project Discription

This project is the final project of the UC Berkeley Data & Analytics Bootcamp. The goal of the project was to develop a Machine Learning application. Our mission was to focus on facial recognition. Our group developed the "Photo Checker Application".

The Photo Checker Application uses Amazona's Rekognition API to recognize famous people in an image that is being uploaded. The Application can detect how many people are in the image and whether the image includes unsafe content.

In addition to that, we included a machine learning feature in the Photo Checker Application that, based on on training data, gives the people in the image a score from 1 to 5 (low to high). People were asked to label the training data i.e. images of Caucasion Males and Females, as well as of Asian Males and Females from 1 to 5, based on the perceived attractiveness of those people.

We trained our Machine Learning Model with this data. Now, any image inputed (celebrity or non-celebrity) will get ranked by our Machine Learning Model.

IMPORTANT

We want to emphazise that our group does not support the ranking of people based on visual apprearance! Our goal was rather to show how a machine learning model can be trained on subjective, labeled data. Since the data labeled is based on subjectivity, the machine learning classification from 1 to 5 is inherently biased.

Machine Learning Models should be designed with caution

Team - The Fantastic Five

Click on our names to get to our GitHub Accounts

• Tanvir Khan
• Paul Pineda
• Nicky Pant
• James Ye
• Fabienne Zumbuehl

Recent news on Facial Recognition

Data

Our Data has been retrieved from the SCUT-FBP5500-Database

Find the link to the Github repo here:

SCUT-FBP5500-Database-Release

About the Data

• 60 raters labeled the data with a "Beauty Score" from 1 to 5
• The distribution of the scores is a Normal Distribution
• People in the images that were labeled were Causasion Male/Female and Asian Male/Female
• Caucation and Asian Females got the highest rating, followed by Causasion Male and Asian Males