Face Recognition using Artificial Neural Network www.ppttopics.com
Contents Problem specification Motivation Design Work done Results Future work Demonstration References www.ppttopics.com
Problem Specification To develop a face recognition system that: Takes a face image of a person as an input. Compares the face image of a person with the existing face images that are already stored in the database. Reports whether it is identified or not. www.ppttopics.com
Motivation Identity fraud is becoming a major concern for all the governments around the globe Reliable methods of biometric personal identification exists ,but these methods rely on the cooperation of the participants neural networks are good tool for classification purposes www.ppttopics.com
Design Image Sampling Karhunen Loeve (KL) Transform Multilayer Perceptron Classification www.ppttopics.com
Image sampling www.ppttopics.com
KL Transform To reduce the dimensions of the image vector Based on eigen values and corresponding eigen vectors. www.ppttopics.com
Multilayer perceptron www.ppttopics.com
Training a neural network We train our neural network with a large sample of images. We wish to find the collection of weights that minimizes || TNET - TACTUAL || . www.ppttopics.com
Testing After training is complete then the system as a whole is ready to be used for recognizing any given image. Testing image is used as an input to our system, the output of the system is compared against the values stored in the database. System reports whether a match or mismatch. www.ppttopics.com
Work Done Main concern in the project: Face recognition and not face detection. Database of preprocessed images taken CMU AMP Face Expression Database contains 975 images of 13 subjects (75 images of each person) ‘bmp’ format with slightly varying poses, expressions etc converted into ‘pgm’ format using GIMP Separate java classes for K L transform Multilayer Perceptron (MLP) Training the MLP www.ppttopics.com
Package named JAMA (Java matrix) used Contains matrix operations like covariance, inverse, transpose etc. Coding done in java. Reasons being: To make application platform independent Java’s ability to handle large numbers Object oriented: to model real life situations www.ppttopics.com
Neural net features: Working Number of input layer neurons: Number of Eigenvalues Number of hidden layers: 1 Number of hidden layer neurons: 24(can be changed) Number of output layer neurons: total number of subjects Output given by neurons: 0 or 1 Working Training done with training images Validation done for the test images Appropriate message generated if subject is identified or not identified www.ppttopics.com
RESULTS Different permutations tried for : Hidden layer neurons Output neurons Form of outputs Training cycles Learning rate Done to bring error in an acceptable range www.ppttopics.com
Satisfactory results obtained for following combination : Input neurons : selected Eigens Hidden neurons : 24(can be changed) Output neurons: total number of different subjects Training cycles: 100000 Learning Rate: 0.3 Error obtained: 2.42E-4 The system identified the subjects presented during training For subjects not given during training : System refused to identify www.ppttopics.com
FUTURE WORK Face detection can be implemented Processing of image can be incorporated Output of unidentified persons can be stored for future reference Ensemble of MLPs can be implemented Incremental learning can be implemented www.ppttopics.com
The mean Image www.ppttopics.com
DEMO www.ppttopics.com
After training www.ppttopics.com
Selecting Image www.ppttopics.com
Match Found www.ppttopics.com
No Match www.ppttopics.com
References [1] Steve Lawrence, C. Lee Giles, Ah Chung Tsoi, Andrew D. Back, Face Recognition: A Hybrid Neural Network Approach, Technical Report, UMIACS-TR-96-16 and CS-TR-3608, Institute for Advanced Computer Studies, University of Maryland, 1996. [2] Wendy S. Yambor, Analysis of PCA-based and Fisher discriminant-based image recognition algorithms, Technical Report CS-00-103, Computer Science Department, Colorado State University, July 2000. [3] Stuart Russel, Peter Norvig, Artificial Intelligence: A Modern Approach, Pearson Education, 2nd Edition. www.ppttopics.com
[4] Matthew A. Turk, Alex P. Pentland, Face Recognition Using Eigenfaces, Vision and Modeling Group, The Media Laboratory, Massachusetts Institute of Technology, 1991. [5] W. Zhao, R. Chellappa, A. Rosenfeld, P.J. Phillips, Face Recognition: A Literature Survey, ACM Computing Surveys, 2003, pp. 399-458. [6] T. De Bie, N. Cristianini, R. Rosipal, Eigenproblems in Pattern Recognition, Handbook of Computational Geometry for Pattern Recognition, Computer Vision, Neurocomputing and Robotics, E. Bayro-Corrochano (editor), Springer-Verlag, Heidelberg, April 2004. [7] Bai-Bo Zhang, Chang-Shui Zhang, Lower Bounds Estimation to KL Transform in Face Representation and Recognition, Proceedings of the First International Conference on Machine Learning and Cybernetics, Beijing, 4-5 November 2002. www.ppttopics.com
[8] An Introduction to Linear Algebra, : http://www. cs. princeton [9] John Heaton ,An Introduction to Neural Networks in Java, http://www.samspublishing.com [10] H.M. Deitel, P.J. Deitel, Java How to Program, Pearson Education,5th Edition www.ppttopics.com