Home > Events > Oral Candidacy - Sandipan Banerjee

Oral Candidacy - Sandipan Banerjee

Start: 5/15/2017 at 2:00PM
End: 5/15/2017 at 5:00PM
Location: 315 Stinson Remick
Attendees: Faculty and students are welcome to attend the presentation portion of the defense.
Add to calendar:
iCal vCal

Sandipan Banerjee

Oral Candidacy

May 15, 2017                       2:00 pm              315 Stinson Remick

Advisers:  Dr. Patrick Flynn and Dr. Kevin Bowyer

Committee:

Dr. Domingo Mery       Dr. Walter Scheirer        Dr. Chaoli Wang       

Title

Exploring the Effects of Frontalization & Data Synthesis on Face Recognition

Abstract

Automatic face recognition performance has improved remarkably in the last decade. Much of this success can be attributed to the development of deep learning techniques like convolutional neural networks (CNNs). But the training process for CNNs requires a large amount of clean and correctly labelled data. If a CNN is intended to work with non-frontal face images, should this training data be diverse in terms of facial poses, or should face images be frontalized as a pre-processing step? To answer this question we evaluate a set of popular facial landmarking and pose frontalization algorithms to understand their effect on facial recognition performance. We also introduce a new landmarking and frontalization scheme that operates over a single image without the need for a subject-specific 3D model, and perform a comparative analysis between the new scheme and other methods in the literature. 

Secondly, we propose a novel face synthesis method for augmentation of existing face image datasets. An augmented dataset reduces overfitting, which in turn, can enhance the face representation capability of a CNN. Our method, starting off with actual face images from an existing dataset, can generate an arbitrarily large number of synthetic images of real and synthetic identities. Thus a face dataset can be expanded both in terms of the number of identities represented (breadth) and the number of images per identity (depth) without the identity-labeling and privacy complications that come from downloading images from the web. Experiments show the synthetic images generated by our method to be unique and training a CNN with face data augmented using this method to boost its recognition performance.