Year: 2013
Type: dissertation
Abstract: Face recognition is a central issue in the fields of Computer Vision and Biometrics, with application in security, robotics, human-computer interfaces, digital cameras and entertainment. Despite the large effort dedicated to improving facial recognition algorithms most of them fail in produce good results in real time applications under different lighting conditions, pose, and facial expression. \nIn recent surveys of face recognition techniques, pose variation has been identified as one of the prominent unsolved problems in the research of face recognition and it is an area of growing interest in the computer vision community. \nThis work presents the development of a pose estimation system, based on convolutional neural networks. The algorithm achieves a recognition rate of 90.1% on the public access database FERET, and 88.6% on the FEI database. \nAdditionally it is shown its use as a complement to a recognition system based on Eigenfaces (PCA) where it allows an increase in the recognition rate from 40.5% to 61.4% over images of faces in different poses taken from the FERET database.
Authors: Méndez Silva, Paul Esteban
Language: Spanish
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2228982388