A Proposal Defense On Deep Residual Network For Face Recognition Presented By SAGAR MISHRA MECE

Slides:



Advertisements
Similar presentations
Face Recognition: A Convolutional Neural Network Approach
Advertisements

ImageNet Classification with Deep Convolutional Neural Networks
Face Recognition Using Neural Networks Presented By: Hadis Mohseni Leila Taghavi Atefeh Mirsafian.
Convolutional Neural Networks for Image Processing with Applications in Mobile Robotics By, Sruthi Moola.
COMPARISON OF IMAGE ANALYSIS FOR THAI HANDWRITTEN CHARACTER RECOGNITION Olarik Surinta, chatklaw Jareanpon Department of Management Information System.
Building high-level features using large-scale unsupervised learning Anh Nguyen, Bay-yuan Hsu CS290D – Data Mining (Spring 2014) University of California,
Object detection, deep learning, and R-CNNs
Convolutional Neural Network
Deep Residual Learning for Image Recognition
Deep Learning Overview Sources: workshop-tutorial-final.pdf
Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition arXiv: v4 [cs.CV(CVPR)] 23 Apr 2015 Kaiming He, Xiangyu Zhang, Shaoqing.
Convolutional Neural Networks
Big data classification using neural network
Deep Residual Learning for Image Recognition
Deep Residual Networks
Demo.
Convolutional Neural Network
Deep Feedforward Networks
The Relationship between Deep Learning and Brain Function
CS 6501: 3D Reconstruction and Understanding Convolutional Neural Networks Connelly Barnes.
Machine Learning for Big Data
Summary of “Efficient Deep Learning for Stereo Matching”
Deep Learning Amin Sobhani.
Data Mining, Neural Network and Genetic Programming
Data Mining, Neural Network and Genetic Programming
Computer Science and Engineering, Seoul National University
Convolutional Neural Fabrics by Shreyas Saxena, Jakob Verbeek
The Problem: Classification
Article Review Todd Hricik.
Applications of Deep Learning and how to get started with implementation of deep learning Presentation By : Manaswi Advisor : Dr.Chinmay.
Inception and Residual Architecture in Deep Convolutional Networks
Intelligent Information System Lab
Natural Language Processing of Knee MRI Reports
Classification of Hand-Written Digits Using Scattering Convolutional Network Dongmian Zou Advisor: Professor Radu Balan.
ECE 6504 Deep Learning for Perception
Deep Belief Networks Psychology 209 February 22, 2013.
Machine Learning: The Connectionist
Deep Residual Learning for Image Recognition
R-CNN region By Ilia Iofedov 11/11/2018 BGU, DNN course 2016.
State-of-the-art face recognition systems
Introduction to Neural Networks
Face Recognition with Deep Learning Method
Image Classification.
A Comparative Study of Convolutional Neural Network Models with Rosenblatt’s Brain Model Abu Kamruzzaman, Atik Khatri , Milind Ikke, Damiano Mastrandrea,
Two-Stream Convolutional Networks for Action Recognition in Videos
CS 4501: Introduction to Computer Vision Training Neural Networks II
Deep learning Introduction Classes of Deep Learning Networks
Object Classification through Deconvolutional Neural Networks
Very Deep Convolutional Networks for Large-Scale Image Recognition
Smart Robots, Drones, IoT
network of simple neuron-like computing elements
LECTURE 35: Introduction to EEG Processing
Neural Networks Geoff Hulten.
Deep Learning for Non-Linear Control
LECTURE 33: Alternative OPTIMIZERS
Visualizing and Understanding Convolutional Networks
Object Tracking: Comparison of
Analysis of Trained CNN (Receptive Field & Weights of Network)
Deep Learning Some slides are from Prof. Andrew Ng of Stanford.
Problems with CNNs and recent innovations 2/13/19
Heterogeneous convolutional neural networks for visual recognition
Face Recognition: A Convolutional Neural Network Approach
Course Recap and What’s Next?
Department of Computer Science Ben-Gurion University of the Negev
Deep Object Co-Segmentation
Natalie Lang Tomer Malach
CS295: Modern Systems: Application Case Study Neural Network Accelerator Sang-Woo Jun Spring 2019 Many slides adapted from Hyoukjun Kwon‘s Gatech “Designing.
VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION
Image recognition.
End-to-End Facial Alignment and Recognition
Presentation transcript:

A Proposal Defense On Deep Residual Network For Face Recognition Presented By SAGAR MISHRA MECE 15945

Face Recognition  Faces represent complex, multidimensional, meaningful visual stimuli and developing a computational model for face recognition is difficult.  It is a very active research topic because of its application in human-robot interaction, human- machine interfaces, driving safety, Security.  Despite of the significant improvements, face recognition is still a challenging problem that wait for more and more accurate algorithms.

WHY DEEP RESIDUAL NETWORK ??

Problem Statement  The deeper the network can cover more complex problems  This is supposed to enhance the accuracy of the network.  However, training the deeper network is more difficult because of two problem  Vanishing Gradients : neuron can “die” in the training and become useless. This can cause information loss  Optimization Difficulty : weights, biases increases due to increasing depth, training the network becomes very difficult.

Consequences of more deeper network  When the number of layer increased than, there is increment in training error and test error  Kaming He el. a [2] present the increment of layers in plain Convolutional network from 20 t0 56 and shows the increment of errors.

Deep residual Learning  Plain net  H(x) is any desired mapping, hope the 2 weight layer fit H(x)  Residual nets  H(x)=F(x)+x, F(x) is residual mapping

Research Objective  To design a Face recognition system using Deep Residual Network  To compare the performance of this network with the previous counterparts VGG-Face and best result of paper "Deeply learned face representations are sparse, selective, and robust“  The Face recognition system will decrease training period and training error using deep residual network.  The comparison is based on following parameters  Accuracy  Training period  Training error

System Model

Face Detection  Detecting face in the set of images is the first step  In our model we are going to use Histograms of Oriented Gradients (HOG)  Steps for HOG transform  Make image black and white  For every single pixel look at the pixel that directly surround it.  Find the direction in which pixel gets darker  Every pixel is replaced with arrow called gradients

Preprocessing  Faces can be aligned in different ways  To account for this face landmark estimation algorithm invented in 2014 by Vahid Kazemi and Josephine Sullivan [5] will be used in our model.  The basic idea is to identify 68 landmark points that exist in every face  Finally Affine Transformation will be used to center the nose, eyes and mouth.

Feature Extraction  Residual network will be used to train with available dataset and extract features for individual image  residual network with depth of 19 layers will be trained. VGG-19 net will also be trained as a contrast for comparison purpose.

Deep Residual Block(ResBlock)

Convolution Neural Network  Consists of input, output and hidden layers  Hidden layers are Convolutional, Pooling and Fully connected  Convolutional Layer  Performs convolution and send output to next layer  Consists of filters  Filters are convoluted across length and width of input image to generate feature map  Its is simply dot product

Convolution Neural Network  Pooling Layer  Also called down sampling  Combines output of neuron clusters at one layer into a single neuron in next layer  Max pooling uses maximum value from each of cluster of neuron.  Fully connected layer  Traditional Multi Layer Perceptron acts as output layer  Connects every neuron in one layer to every neuron in another layer

Convolutional Neural Network

Rectified Linerar Unit (ReLU)  element wise operation (applied per pixel) and replaces all negative pixel values in the feature map by zero

Face Recognition  In the final step featured extracted from the image under test must be compared with the features stored in the face database to identify the face.

Training the network  The entire network will be trained with Stochastic Gradient Descent (SGD) with backpropagation.

Training Dataset  Dataset that will be used is LFW (Labeled Face in the Wild)  Consists of faces  Each face labeled with name of the person

Proposed Model

Performance Metric  The result of this model will be comparison with VGG-Face and best result of the paper [6]. Our model will be compared with previous models against accuracy, training error, training period.

Tools  The tools, programming language and software’s that will be used in this thesis work are listed below:  Python programming Language  Eclipse (pycharm)  Caffe/ Keras platform

Expected Output  Implementation of deep residual network for face recognition  The result of the residual network will be evaluated and compared with other best models against accuracy, training error, training period

Schedule TasksMonth/Year(2017/2018) SepOctNovDecJanFeb Literature Review Proposal Defense System Design And Coding Mid-Term Defense Final Submission of Thesis Documentation of Thesis Research and Experiments

References  [1] X. Z. S. R. He, Kaiming and J. Sun, “Deep residual learning for image recog-nition,”ICCV,  [2] A. D. B. Steve Lawrence, “Face recognition: A convolutional neural networkapproach,”IEEE, 1997  [3] J. D. S. K. Y. Jia, E. Shelhamer, “Caffe: Convolutional architecture for fastfeature embedding,”arXiv: , 2014  [4] Z. C. Xiu Li, “Deep redisual network for plank classification,” ICCV,  [5] A. Z. K. Simonyan, “Very deep convolutional networks for large- scale image recognition,”ICLR, , 6  [6] X. W. Y. Sun and X. Tang, “Deeply learned face representations are sparse, selective, and robust,”CoRR,abs/ , 2014  [7] S. I. Krizhevsky, A. and G. E. Hinton, “Imagenet classification with deep convolutional neural networks.” NIPS, 2012  [8] H. W. DMasaki Nakada and D. Terzopoulos, “Acfr:active face recognition using convolutional neural network,” 2017