Scribd Logo
Upload

Welcome to Scribd!

  • 37 views

    Hizami CHAPTER 1-Feedbacks

    Uploaded by

    Mohd Hizami Ab Halim
    This document outlines a project to use convolutional neural networks and compound scaling methods to classify offline signature images. It will investigate ResNet and EfficientNet architectures, applying various scaled versions to a dataset of 54,000 signatures from 1,000 users. The objectives are to develop and evaluate CNN models for signature classification, analyzing their accuracy and error rates. Key questions include exploring deep learning approaches for signature verification using compound scaling, and how EfficientNet architectures perform on this task. The project scope involves studying ResNet and EfficientNet, using Python for model training and testing on the GPDS Synthetic Signatures Database.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    Hizami CHAPTER 1-Feedbacks

    Uploaded by

    Mohd Hizami Ab Halim
    37 views5 pages
    This document outlines a project to use convolutional neural networks and compound scaling methods to classify offline signature images. It will investigate ResNet and EfficientNet architectures, applying various scaled versions to a dataset of 54,000 signatures from 1,000 users. The objectives are to develop and evaluate CNN models for signature classification, analyzing their accuracy and error rates. Key questions include exploring deep learning approaches for signature verification using compound scaling, and how EfficientNet architectures perform on this task. The project scope involves studying ResNet and EfficientNet, using Python for model training and testing on the GPDS Synthetic Signatures Database.

    Original Description:

    School based

    Original Title

    Hizami CHAPTER 1-feedbacks

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document outlines a project to use convolutional neural networks and compound scaling methods to classify offline signature images. It will investigate ResNet and EfficientNet architectures, applying various scaled versions to a dataset of 54,000 signatures from 1,000 users. The objectives are to develop and evaluate CNN models for signature classification, analyzing their accuracy and error rates. Key questions include exploring deep learning approaches for signature verification using compound scaling, and how EfficientNet architectures perform on this task. The project scope involves studying ResNet and EfficientNet, using Python for model training and testing on the GPDS Synthetic Signatures Database.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    37 views5 pages

    Hizami CHAPTER 1-Feedbacks

    Uploaded by

    Mohd Hizami Ab Halim
    This document outlines a project to use convolutional neural networks and compound scaling methods to classify offline signature images. It will investigate ResNet and EfficientNet architectures, applying various scaled versions to a dataset of 54,000 signatures from 1,000 users. The objectives are to develop and evaluate CNN models for signature classification, analyzing their accuracy and error rates. Key questions include exploring deep learning approaches for signature verification using compound scaling, and how EfficientNet architectures perform on this task. The project scope involves studying ResNet and EfficientNet, using Python for model training and testing on the GPDS Synthetic Signatures Database.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 5

    Working title?

    CHAPTER 1 Post Vaccine symptom


    using deep learning /
    Fake Signature Recognation
    INTRODUCTION using deep learningSV(s):

    1.1 Introduction

    Today, the use of biometrics has become a necessity and is widely exploited by most
    organizations especially in signature verification. Biometrics refers to the method of identity
    recognition through the biological and behavioural characteristics inherent in everyone.
    Biometric systems are proven to have a high level of accuracy as everyone has unique
    biological and behavioural characteristics. Among the applications that use of biometrics are
    the areas of security, forensics and attendance systems. Two types of signatures verification
    are online and offline. For the online signature verification, it captured during writing
    process, therefore can get active information of signature, such as the stroke order, the
    pressure applied and the speed. For the offline signature verification, it analyses a stationary
    image.

    Convolutional Neural Network (CNN) has made a powerful tool for image
    classification tasks. In machine learning, a Convolutional Neural Network (CNN) is the class
    of deep learning, feeding the artificial neural networks and commonly applied for analysing
    visual image. CNN use multilayer perceptron to require minimal pre-processing. These
    biologically inspired computational models can outperform the previous forms of artificial
    intelligence in common machine learning tasks.
    1.2 Background Of The Problem

    ResNet has won 1st price in all ImageNet Large Scale Visual Recognition Challenge
    (ILSVRC) 2015 and Microsoft Common Objects in Context (MS COCO) 2015 based
    techniques show state-of-the-art accuracy on the ImageNet task (Kaiming He et. al.).
    (Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun, 2015) recently published a paper on
    the ImageNet dataset and the state-of-the-art accuracies achieved during last few years. The
    paper shows the evolution of deep learning techniques overtime on this challenge from 2012.
    ResNet-152 showed only 3.57% error, which is better than human error for this task at 5%.

    EfficientNet is a convolutional neural network architecture and scaling method that


    uniformly scales all dimensions of depth/width/resolution using a compound coefficient.
    Unlike conventional practice that arbitrary scales these factors, the EfficientNet scaling
    method uniformly scales network width, depth, and resolution with a set of fixed scaling
    coefficients.

    This project aim is to use the compound scaling method to scale the dimensions of
    the network. The applied grid search strategy to find the relationship between the different
    scaling dimensions of the baseline network under a fixed resource constraint. This project is
    organized as follows: first it states the problem, then discusses the details of dataset that are
    evaluated. Then it describes the procedure of selected architectures and finally discuss the
    obtained results.

    1.3 Problem Statement

    This project investigates the compound scaling method to scale the dimensions of the
    network which CNN architecture for classification of offline signature is most accurate and
    performs well. Performance of several architectures will be studied from literature and their
    analysis would be compared to the RestNet and EfficientNet which would be used for the
    mentioned problem. Using this strategy, they could find the appropriate scaling coefficients
    for each of the dimensions to be scaled-up. Using these coefficients, the baseline network was
    scaled by the desired size.

    1.4 Project Aim

    This project will study all aspects of a Convolutional Neural Networks and the
    architecture layouts of RestNet and EfficentNet were then be applied these on GPDS
    Synthetic Signature Database dataset of signatures using Python programming language and
    its libraries and then compare the obtained results will be compared to the results that are
    available in literature.

    1.5 Project Objectives

    The objectives of this project are:

    (a) To study and investigate various deep learning algorithms for RestNet and
    EfficientNet.

    (b) To develop compound scaling method to scale the dimensions of the network for
    classifications of signature images.

    (c) To train, test and validate CNN architectures for signature classification.

    (d) To evaluate and analyse the classification accuracy performances, and Equal Error
    Rate (EER) for GPDS Synthetic Signature Database (1000 users) using compound
    scaling method architectures.
    1.6 Project Questions

    The main questions that this project intents to answer are:

    (a) What are deep learning approaches on compound scaling method in signature
    verification?

    (b) How Convolutional Neural Networks (CNN) are designed using EfficientNet versions
    signature classification?

    (c) How well would these architecture variations will perform for signature classification
    for 54000 samples from 1000 users?

    1.7 Project Scope

    The scope of this project will be:

    (a) Focus on CNN algorithm RestNet and EfficientNet architecture details.

    (b) Python programming language is used for training and testing of the
    Convolutional Neural Networks.

    (c) GPDS Synthetic Signatures Database (1000 users) is used as dataset for the
    compounding the scaling method.
    1.8 Summary

    This chapter introduces Deep Convolutional Neural Networks and its potential
    applications in image classification and signatures classification. The background of this
    project, and the problem this project tends to answer are stated including aims and objectives.
    In the next chapter (Chapter 2) discusses on the literature needed to study regarding this
    project. It contained the discussion about the scaling method used for signature verification,
    the EfficientNet and RestNet with be the components and various architectures that be used
    in CNN. In the next chapter (Chapter 3) is the methodology, the chapter that discusses the
    proposed method to tackle the problem. Various phases involved in methodologies. This
    project follows an operational framework with four phases including Study, Investigation,
    Design and Implementation. The next chapter (Chapter 4) discusses on the results and
    analysis of vii the tables and graphs obtained in the proposed work. And evaluated how each
    architecture performed on GPDS Synthetic Signature Database. Also, the chapter double
    checks if this project has achieved its defined objectives, and whether the problem question
    has been answered. Furthermore, it concludes all the work of this project and states future
    enhancements and future work for this project. The last chapter (Chapter 5) concludes the
    results, achievements of this project and mentions novelty of this comparative analysis
    project. Furthermore, future improvement suggestions and future work are discussed.

    You might also like