Thesis presentation
Download as PPTX, PDF•1 like•95 views
This document proposes a local receptive fields based extreme learning machine (ELM-LRF) for face recognition. ELM-LRF introduces local receptive fields to the input layer of an ELM for locally connected neural networks. It is tested on three face datasets: Caltech, CBCL, and UFI. ELM-LRF achieves high testing accuracies of 98.15%, 98.34%, and 66.11% respectively, outperforming other methods. The key advantages of ELM-LRF are that it reduces training time, provides fast results with no risk of getting stuck in local minima like backpropagation algorithms.
![CBCL face dataset(cont.)
Methods Testing Accuracy (%)
ELM-LRF 98.34
Pose invariant 95.40
Original C2 features [14] 87.05
MPCALDA [15] 88.53](https://image.slidesharecdn.com/thesispresentation-180215115720/85/Thesis-presentation-26-320.jpg)
Thesis presentation
- 1. Local Receptive Fields Based Extreme Learning Machine For Face Recognition By Aras M. Ismael 14 February 2018
- 2. Outline • Introduction • Brief introduction to the Humans Brain • How humans brain work • Biological inspiration • Artificial Neural network • Neural network modules • Type of Neurons • Face recognition algorithms • Extreme learning machine • Testing • Conclusion
- 3. Introduction As the necessity for higher levels of security rises, technology is bound to swell to fulfill these needs. Any new creation, enterprise, or development should be uncomplicated and acceptable for end users in order to spread worldwide. This strong demand for user-friendly systems which can secure our assets and protect our privacy without losing our identity in a sea of numbers.
- 4. Introduction (cont.) Biometrics is the emerging area of bioengineering; it is the automated method of recognizing person based on a physiological or behavioral characteristic. There exist several biometric systems such as signature, finger prints, voice…etc
- 5. Brief introduction to Humans Brain • The knowledge concerning how a human brain works perfectly has remained unknown over time. • However, much is known about information is processed. The function of neurons is to gather signals emanating from other neurons through dendrites
- 6. How humans brain work
- 7. Biological inspiration synapses axon dendrites The information transmission happens at the synapses. electrochemical stimulation received from other neural cells to the cell body by dendrites.
- 8. Biological inspiration The spikes travelling along the axon of the pre-synaptic neuron trigger the release of neurotransmitter substances at the synapse. The neurotransmitters cause excitation or inhibition in the dendrite of the post-synaptic neuron. The contribution of the signals depends on the strength of the synaptic connection.
- 9. Artificial neurons Neurons work by processing information. They receive and provide information in form of spikes.
- 10. Neural networks
- 11. Neural network modules(cont.) Multi layer Neural network
- 12. Back propagation
- 13. Back propagation problem(cont.) Delta(learning rule): which it implements weight changes and The learning rule in a multilayer perceptron is not guaranteed to produce convergence, and it is possible for the network to fall into a situation (the so called local minima) in which it is unable to learn the correct output. Isolation
- 14. Applications of neural networks Character Recognition - The idea of character recognition has become very important as handheld devices like the Palm Pilot are becoming increasingly popular. Neural networks can be used to recognize handwritten characters. • Image Compression - Neural networks can receive and process vast amounts of information at once, making them useful in image compression. With the Internet explosion and more sites using more images on their sites, using neural networks for image compression is worth a look.
- 15. Applications of neural networks(cont.) • Stock Market Prediction - The day-to-day business of the stock market is extremely complicated. • Medicine, Electronic Nose, Security, and Loan Applications. • Health usage • Face recognition applications
- 16. Neural network in face recognition
- 17. Local Receptive Fields Based Extreme Learning Machine For Face Recognition
- 18. Extreme learning machine • ELM was proposed for single-hidden layer feedforward neural networks (SLFNs). It is very different from conventional neural network learning algorithms. It randomly chooses the parameters of hidden nodes and analytically determines the output weights. Thus the training is extremely fast and efficiently completed without time- consuming iterations.
- 19. Extreme learning machine(cont) • The name "extreme learning machine" (ELM) found by Guang-Bin Huang. • these models are able to produce good generalization performance and learn thousands of times faster than networks trained using backpropagation.
- 20. Extreme learning machine (cont.) the input weights and hidden biases are randomly generated, and the output weights are analytically determined by regularized least square method, It presents better accuracy and high efficiency, in various applications such as system modelling, biomedical analysis, power systems, etc.
- 21. Local respective fields LRF-ELM introduces local receptive field to the input layer, thus obtaining a locally connected ELM. The hidden layers of LRF-ELM consists of a convolution layer and a pooling layer. They are composed of several feature maps. The input weights between input and convolution layers are first randomly generated according to some continuous probability distribution and then orthogonalized in order to obtain a more complete set of features
- 22. ELM-LRF implementation using the following datasets • Caltech face dataset • Cbcl face dataset • UFI face dataset
- 23. Caltech face dataset The dataset has 10,524 human appearances of different resolutions and in various settings The Caltech confront dataset has 450 images of 27 people. Every ha diverse helping qualities and a size of 64×64 pixels.
- 24. Caltech face dataset(cont.) • Based on ELM-LRF testing accuracy in the given dataset can see that ELM-LRF has more advantage over face recognition compare to other methods test in this dataset. Method Testing Accuracy (%) NR MODEL 32.36 SCSPM 82.83 TSR 37.45 SPARSE BASED NN 92.91 ELM-LRF 98.15
- 25. CBCL face dataset The training set comprises of 6,977 images (2,429 faces and 4,548 nonfaces), and the test set comprises of 24,045 images (472 countenances and 23,573 nonfaces). The MIT-CBCL dataset contains 2,000 face pictures. The calculation utilized 150 images for training (15 images for each class with right, left, and frontal perspectives).
- 26. CBCL face dataset(cont.) Methods Testing Accuracy (%) ELM-LRF 98.34 Pose invariant 95.40 Original C2 features [14] 87.05 MPCALDA [15] 88.53
- 27. UFI dataset This dataset contain images of 605 people. The images are cropped to a size of 128 x 128 pixels.
- 28. UFI face dataset(cont.) • In testing accuracy by using UFI dataset can be seen that the testing accuracy of the proposed ELM-LRF method is an enhancement on LBPHS is 55.44 %, LDPHS, and FS-LBP, although it achieved the face size also significantly differs and the faces are not localized after that it slightly lower test accuracy than that of POEMHS which reduced 12.06% difference. Method Testing Accuracy (%) LBPHS 55.04 LDPHS 50.25 POEMHS 67.11 FS-LBP 63.31 ELM-LRF 66.11
- 30. Conclusion Reduced training time Fast result No isolation in this method The outcome can be found in one iteration
- 31. Conclusion (cont.) based on our result by using ELM-LRF in face recognition the proposed method can have more advantage in face recognition systems because it will set the best weight for the given input Because the input has local connection with the desired output, it can set the best weight for the given output. Based on our result , we can say that the ELM-LRF can pass back- propagation problems which it can not find its weight easily.