A Comparative Study on Identical Face Classification using Machine Learning

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

A Comparative Study on Identical Face Classification using Machine Learning

1Student of Masters of Engineering, Ahmedabad, Dept of Computer Engineering, L.J. Institute of Engineering & Technology, Gujarat, India

2Assistant Professor, Ahmedabad, Dept of Computer Engineering & Technology, Gujarat, India ***

Abstract - This paper presents a new identical face classification using Support Vector machine (SVM) classification tree. Automatic face classification from facial image attains good accuracy with large set of training data. While face attribute classification from facial image still remains challengeable. In this paper, we improve the same faces features. Each SVM acts an independent membership/non-membershipclassifierandseveral SVMare combined in a plurality voting scheme that chooses the classificationmade bymorethanthehalfofSVMs.Theglobal systems classify faces by classifying a single feature vector consisting of the gray values of the whole face image. For a good encoding of face efficient reduction of data dimension andstrongseparationofdifferentfaces,respectively.Next,the SVM ensemble is applied to authenticate an input face image whether it is included in the membership group or not. We compare the SVMs based classification with the standard identical approach.

Key Words: Face Classification, Haarcasecade, Wavelet Transform, Support vector Machine, Random Forest, Logistic Regression

1. Introduction

It is well classification that the use of face pictures for personal surveillance, identification and verification is a rapidlyexpandingfieldofstudyinmanyComputersvision application.Someexamplesinthisareaarefacerecognition face action, classification, faces recognition, skin color classification, age estimation, gender recognition and ethnicityrecognition.Faceclassificationhasachievedbetter results according to the research done for nearly three decades. However, similar accuracy of classification could notbegainedfromfacialattributerecognition.

The most effective and powerful classifier for pattern recognitionisfoundinthehumanbrain.Itsamazingpower comes from the fact that it is a dynamic organ engaged in trainingandlearningforaredeterminedamountoftime.In ordertoachievethesameorbetterresults,researcher’skey goalsistoreplicatethisbiologicalandbehavioraltraitofthe human brain in artificial neurons. The classification of human characteristic like identical faces, age gender and ethnicity using facial photographs is one of the main

applicationsofthiswork.Inthisstudy,weaimtodevelopa precise technique for identifying the same faces from facialphotographs.

Identicalfaceclassificationisdoneaccordingtogeometric difference of primary feature in male and female. This algorithmcanclassifythefacialimagesintofacesfeatures. Identical classificationisbasedonthetexturevariation of eyelids,wrinkledensityintheforehead,lipsandcheekarea. Classification is done using SVM and two separate neural networksforidenticalfaceclassification.

1.1 Machine Learning:

Machinelearningisagrowingtechnologywhichenables computers to learn automatically from past data. Machinelearningusesvariousalgorithmsforbuilding mathematical models and making predictions using historicaldataorinformation. Currently,itisbeingused for various tasks such asimage recognition, speech recognition, email filtering, Facebook autotagging, recommendersystem, andmanymore.[2]

1.1.2

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

Fig-3 MachineLearningModel[3]

Machine Learning Techniques:

Supervised Learning:

Itisdefinedbyitsuseoflabelleddatasetstotrain algorithmsthattoclassifydataorpredictoutcomes accurately [4]. Supervised Algorithms are Naïve Bayes,LogisticRegression,SupportVectorMachine, RandomForestandmanymore.

Fig-6: ReinforcementLearningModel[6]

1.1.3 Machine Learning Algorithm:

1. Logistic Regression: Itisoftenusedforclassification andpredictiveanalytics.Logisticregressionestimates theprobabilityofaneventoccurring,suchasvotedor didn’tvote,basedonagivendatasetofindependent variables. Since the outcome is a probability, the dependent variable is bounded between 0 and 1. It usestheactivationfunctioncalledSigmoid.

Fig-7: LogisticRegressionEquation[7]

Sigmoid Curve:

Fig-4: SupervisedLearningModel[4] 

Unsupervised Learning:

Thesealgorithmsdiscoverhiddenpatternsordata groupingswithouttheneedforhumanintervention. Unsupervised Algorithms are Association and Clustering.

Fig-6: SCurveforSigmoid[8]

2. SVM Algorithm:

Fig-5: UnsupervisedLearningModel[5] 

Reinforcement Learning:

It isan area of machine learning concerned with howintelligent agentsought to take actions in an environment in order to maximize the notion of cumulativereward.Machinemakesobservationof rewardsandactsaccordingly.

Support Vector Machine or SVM is one of the most popularSupervisedLearningalgorithms,whichisused for Classification as well as Regression problems. However, primarily, it is used for Classification problems in Machine Learning. The goal of the SVM algorithm is to create the best line or decision boundarythatcansegregaten-dimensionalspaceinto classessothatwecaneasilyputthenewdatapointin thecorrectcategoryinthefuture.Thisbestdecision boundary is called a hyperplane. SVM chooses the extreme points/vectors that help in creating the

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page824

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

hyperplane.Theseextremecasesarecalledassupport vectors, and hence algorithm is termed as Support VectorMachine.Considerthebelowdiagraminwhich there are two different categories that are classified usingadecisionboundaryorhyperplane[9]:

Fig-9: SVMWorking[10]

Fig-10: SVMEquations[11]

3. Random Forest Classifier:

ItcanbeusedforbothClassificationandRegression problemsinML.Itisbasedontheconceptofensemble learning,which is a process ofcombining multiple classifierstosolveacomplexproblemandtoimprove the performance of the model. Random Forest is a classifierthatcontainsanumberofdecisiontreeson various subsets of the given dataset and takes the average to improve the predictive accuracy of that dataset.Instead of relying on one decision tree, the randomforesttakesthepredictionfromeachtreeand based on the majority votes of predictions, and it predictsthefinaloutput.Thegreaternumberoftrees intheforestleadstohigheraccuracyandpreventsthe problemofoverfitting.[12]

The below diagram explains the working of the RandomForestalgorithm:

Fig-11: RandomForestWorking[12]

1.2 Literature Survey:

Reference [13] has proposed a novel method using supportvectormachine,MultilayerPerceptionandKNN as representation of facial expression classification is implemented as a classifier for expression SVM deals with the original image by preserving the useful informationandreducingthefeaturevector’sdimension of data. The authors proposed the method of SVM for extractingthequantityofthefacefeaturethathasgrate abilityinimprovingthegeneralizationperformanceand the accuracy of the feature vector. According to this proposedmethodistheimprovedmethodofSVMthat work through changing the feature vectors for producingthetransformationofhighdimensionaldata intothelowdimensionaldata.Theauthorsfoundthat theclassifierofSVM,KNNmethodisveryeffectivefor identifyingtheotherexpression.Theclassificationrate gained 93.89% on database which is applied on eight expressions.

Reference [14] present a CFA system to classify face emotions.Itutilizedthemethodforfeatureextraction andusedmachinelearningalgorithmforclassification which are PCA, LFA, ICA, Naïve and compared their results.Thepresentmethodappliedonimagesequences of subjects. The classification accuracy gained 79.3%, whileusingPCA,81.1%whileusingLFAand95.5%ICA.

Reference [15] in there’s literature work the authors develop system using deep learning methods. It combines the results bases on age and gender Neural Network and tries to predicts the results. In Gender classification, it provides accuracy of 87.06%. In age classification,itgivesaccuracyof79.09%.

Reference[16]inthere’sliteratureworkauthorsusing MachineLearningalgorithmi.e.GaussianMixtureModel (GMM) for face age estimation present promising results.AndHiddenMarkovModel(HMM)supervector torepresentfaceimagepatches.Thedatasetcontaining twodatasetsoffacialimages,onewith4000imagesof 800 females and the others with 4000 images of 800 males.Eachsubsethastheagerangeof0to93.Using

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

GMM they got the accuracy 72.46%, and HMM the accuracy 63.35% in female. Using GMM they got accuracy 58.53% and HMM the accuracy 53.97% in male.

Reference[17]inthere’sLiteratureworkstheauthors develop system using deep learning methods. It combinestheresultsbasesonJointFaceDetectionand Alignment Using Multitask Cascaded Convolutional Networks. It provides accuracy of 95.4% using ConvolutionalNeuralNetwork.

1.3 Data Description:

Dataset is being used which is downloaded from internet,consistsof789imagesfordifferent8persons. Theimagesarehavingdifferentangleoffaces,different expression.

1.4 Proposed System:

A Proposed System uses various algorithms such as Logistic Regression, SVM Classifier, Random Forest Classifier.OurProposedsystemhasseveralphasesinordertoimprovetheAccuracyofresults.

Step 2: The next step is casecade the images using opencv.

Step 3: Inthisstepcroptheallimagesbylength and width offaceusingeyeslidsandlip.

Step 4:Thenextsteptopre-processingthedata and wavelettransformationofimages.

Step 5: In this step compute the results and performance of each algorithm is evaluated to know thebestmodelforclassification.

1.5 Conclusion:

The machine learning algorithm for classification procedure has discussed here. Thus or proposed algorithmhasconcludedthattheoverallperformanceof SVMishightogetaccurateclassificationresultsforIFC. Theappropriateevaluationisneededtodeterminethe classification results of proposed architecture. In our proposed architecture the most relevant feature has beenusedandtestedfortheclassification,besidesthe datasetisveryclearduetonoiseremovethefilters.The irrelevantdatasetshouldbeintroducedandtestedfor theoverall performanceoftheproposedarchitecture. To obtain better efficiency, the SVM method will be validatedforthehugeamountofdataset[18].Machine learningmodelscanbetrainedtoactwellaswehumans doandlearnfromnature.

References:

[1] Aanmeldenofregistrerenomtebekijken. (n.d.). https://www.facebook.com/unsupportedbrowser?igshid=Y mMyMTA2M2Y=

[2] Machine Learning Tutorial | Machine Learning with Python - Javatpoint. (n.d.). www.javatpoint.com. https://www.javatpoint.com/machine-learning

[3] machine learning - Google Zoeken. (n.d.). https://www.google.com/search?q=machine+learning

[4] Supervised Machine learning - Javatpoint. (n.d.). www.javatpoint.com. https://www.javatpoint.com/supervised-machine-learning

[5]Sharma,M.(2021,December11).ABeginnersGuideto Unsupervised Learning - Analytics Vidhya. Medium. https://medium.com/analytics-vidhya/beginners-guide-tounsupervised-learning-76a575c4e942

Fig-10: ProposedModel

Step 1: Thefirststepistoloadtheimages and transform theimagesintothegray color.

[6] Weng, L. (2018, February 19). A (Long) Peek into Reinforcement Learning. Lil’Log. https://lilianweng.github.io/posts/2018-02-19-rloverview/

[7] Logistic Regression in Machine Learning - Javatpoint. (n.d.). www.javatpoint.com. https://www.javatpoint.com/logistic-regression-inmachine-learning

[8] Logistic Regression. (n.d.). https://www.saedsayad.com/logistic_regression.htm

[9]Support VectorMachine (SVM)Algorithm - Javatpoint. (n.d.). www.javatpoint.com. https://www.javatpoint.com/machine-learning-supportvector-machine-algorithm

[10] Team, T. A. I. (2020, June 19). Support Vector Machine   Insights. Towards AI. https://towardsai.net/p/machine-learning/support-vectormachine-an-insight-cdae000758dc

[11]Agarwal,A.(2022,June13).SupportVectorMachine Formulation and Derivation. Medium. https://towardsdatascience.com/support-vector-machineformulation-and-derivation-b146ce89f28

[12] Machine Learning Random Forest AlgorithmJavatpoint. (n.d.). www.javatpoint.com. https://www.javatpoint.com/machine-learning-randomforest-algorithm

[13] Dino, H. I., & Abdulrazzaq, M. B. (2019, April). Facial expression classification based on SVM, KNN and MLP classifiers. In 2019 International Conference on Advanced Science and Engineering (ICOASE) (pp.70-75).IEEE.

[14] Donato, G., Bartlett, M. S., Hager, J. C., Ekman, P., & Sejnowski,T.J. (1999).Classifyingfacialactions. IEEE Transactions on pattern analysis and machine intelligence,21(10),974-989.

[15] Kalansuriya, T. R., & Dharmaratne, A. T. (2013, December). Facial image classification based on age and gender.In 2013InternationalConferenceonAdvancesinICT for Emerging Regions (ICTer) (pp.44-50).IEEE.

[16]Zhuang,X.,Zhou,X.,Hasegawa-Johnson,M.,&Huang,T. (2008, December). Face age estimation using patch-based hidden markov model supervectors. In 2008 19th International Conference on Pattern Recognition (pp. 1-4). IEEE.

[17]Zhang,K.,Zhang,Z.,Li,Z.,&Qiao,Y.(2016).Jointface detection and alignment using multitask cascaded convolutional networks. IEEE signal processing letters, 23(10),1499-1503.

[18]Chen,J.I.Z.,&Hengjinda,P.(2021).Earlypredictionof coronaryarterydisease(CAD)bymachinelearningmethodacomparativestudy. Journal of Artificial Intelligence, 3(01), 17-33

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056 Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page827