sherwyn11 / Synergy-Hackathon-Facial-Recognition

A web-app which performs facial recognition

This project was made for the Synergy Hackathon on 30th August 2019.

The Problem Statment:

Facial Recognition using Machine Learning: Construct a machine learning model that is trained using images of faces of celebrities and tagging features like hair(straight, curly, wavy, long, short, colour, etc), nose, ear, eyes, face shape, etc (features to tag upon can be anything you deem is useful and make the model more accurate). A user screen where the user can select various features of a person they want to look up and the model automatically shows up accurate images that match the description. The prospect of this problem is in criminal recognition and removing redundant sketch work. For the purpose of judging, celebrity pictures should be used. Minimum training set should consist of 1,000 images.

Features:

1. User can select features to fetch required images from the database.
2. An input image will be auto-labelled and stored into the database.
3. Livestreaming from the webcam with live facial feature tagging.
4. A pix2pix implementation by Zaid Alyafeai

Areas to Improve Upon:

1. Using FastAi for our model.
2. Using image hashing techniques to store images.
3. Improving the overall UI.