Evaluation Of Deep Learning Models On Facial Behavior Analysis Of Emotions Detection And Its Applications

The identification of facial behavior analysis and emotions detection(FBAEDA)is an area of research in which the expressions on the face,identify the human emotions.Recent research on facial behavior has become more popular in this era due to the rapid increase of information technology.As the world is fast changing,and we depending on various technologies as a part of human machine interaction.In that,the security of information across the world has been recognized seeking more attention to solve critical issues that can affect the entire world.The study support the idea of developing technology with multiple real-life applications in dealing with facial behavior detection.The understanding of human expression is very important to human computer interaction,and the technology to identify human facial behavior.Detection of facial behavior is essentially bias when predicted by humans.The study deploy a mixture of databases from 2013-Fer database and Africa students of Southwest University of Science and Technology(ASSDB)for fair assessment.This study performs a systematic evaluation of deep learning models such as;ResNet50,AlexNet,and VGG networks on facial behavior analysis of emotions to know which model can perform better in predicting facial behavior with high accuracy.A structured design will be implemented in the process to help achieve the objectives of the study such includes;image database,training and test sets,data pre-processing,feature extraction,support vector machine(SVM)training,model evaluation,prediction,check performance,classifier evaluation,and test image classification Webcam.A model design is created to train the database in predicting various facial behavior.Texture domain extractors along with different techniques for enhancing facial behavior analysis(FBA).Advanced 2D and 3D facial processing techniques like Edge Oriented Histogram(EOH)and Facial Mesh Distance(FMD)will then be fused with a framework to examine the individual facial parts and combined performances of the domains.Using advanced facial orientation and localization methods,the face is then divided into facial sections after these assessments.Deep learning in the context of convolutional neural network(CNN)is also being explored in FBAEDA.It can become an efficient tool for classifying facial behavior.However.The model concept is used to learn the facial parts together,thereby improving over the use of the entire face.Moreover,the study uses a live webcam technology to detect facial expressions which is useful in human society.It can be used on applications such as smart human-computer interaction,biometric security detection,robots and medical psychological behaviors.The results of the experiment shows that,in evaluating deep learning models of facial behavior analysis of emotions detection,Resnet50 predicted the facial expression with a high accuracy which indicates the best network model than AlexNet and VGG networks.