Project Abstract

FRANN explores the topic of face recognition using neural networks, a type of artificial intelligence that attempts to imitate the way a human brain works. Artificial neural networks are composed of neurons that are connected through synapses or weights. Each neuron performs a simple calculation that is a function of the activations of the neurons that are connected to it. Through feedback mechanisms and/or the nonlinear output response of neurons, the network is capable of performing extremely complicated tasks. Key design issues we will be addressing are approximation: is the system capable of accurately approximating the desired relationship; estimation: how much training data will it need; and computation: how should it best use that data to compute its predictions. Current state of the art face recognition technology allows for a recognition accuracy of 95% on more than 1000 frontal mug shot-like images when taken on the same day, our goal is to achieve at least an accuracy of 80% on images taken with changes in lighting, different facial expressions, and pose variations.

Documentation

Fall 2003:

Spring 2004: