A Project Report OnFacial Expression Recognition Using Deep

A Project Report OnFacial Expression Recognition Using Deep Neural NetworkSubmitted in partial fulfillment of the requirement for the 8th semesterBachelor of EngineeringinComputer Science & EngineeringVISVESVARAYA TECHNOLOGICAL UNIVERSITY, BELGAUMSubmitted bySHIVANI DUTT [1DS15CS099]MEENAKSHI BHAT [1DS15CS057]SHALINI [1DS15CS093]SHWETA SANJAY SAVALGI [1DS15CS105]Under the guidance ofDr.Shubha BhatAssociate Professor,CSE,DSCE2018-2019Department of Computer Science & Engineering DAYANANDA SAGACOLLEGE OF ENGINEERING BANGALORE – 56007VISVESVARAYA TECHNOLOGICALUNIVERSITYDAYANANDA SAGAR COLLEGE OF ENGINEERINGShavige Malleshwara Hills, Kumaraswamy Layout, Bangalore – 560078Department of Computer Science & EngineeringCERTIFICATEThis is to certify that the project entitled “Facial Expression Recognition UsiDeep Neural Networks” is a bonafide work carried out by Shivani D[1DS15CS099], Meenakshi Bhat [1DS15CS057], Shalini [1DS15CS093] aShwetha Sanjay Savalgi[1DS15CS105] in partial fulfillment of 8th semestBachelor of Engineering in Computer Science and Engineering under VisvesvaraTechnological University, Belgaum during the year 2018-19.

Dr. Shubha Bhat Dr. Ramesh Babu Dr. C P S Prakash(Internal Guide) Vice Principal Principal, DSCEAssociate professor , Head of Department, Department of Department CSE, DSCECSE, DSCESignature:………………….

. Signature:………………….. Signature:……………… Name of the Examiners: Signature with date:1……………………………………. ……………………………… 2……………………………………. ………………………………ABSTRACTAutomated Facial Expression Recognition (FER) has remained a challenging problem. Present approaches lack generalizability when applied to unseen images or those that arecaptured in wild setting. In this paper we are using deep learning to detect facial expression. We are focusing on expression like neutral, happy, surprise, sad and angry. Weimplemented a general convolutional neural network (CNN) building framework fordesigning real-time CNNs. We have used FER-2013 emotion dataset and IMDB genderdataset. Haar feature-based cascade classifier has been used to detect frontal face.

Get quality help now

Marrie pro writer

Verified

Proficient in: Artificial Intelligence

5 (204)

“ She followed all my directions. It was really easy to contact her and respond very fast as well. ”

+84 relevant experts are online

Hire writer

Initiallywe have processed the images followed by training our model in Keras. The training setconsists of 35,888 examples. Our model take bounded face (48*48 pixels) as input. Wehave also trained our model to detect gender i.e. Male or Female. We conductedcomprehensive experiments on seven well- known facial expression databases (viz. MultiPIE, MMI, CK+, DISFA, FERA, SFEW, and FER2013) and obtain re- sults which aresignificantly better than, or comparable to, traditional convolutional neural networks orother state-of- the-art methods in both accuracy and learning time.Our model can be used todetect gender and emotion on static as well as real time images. Multiple faces can bedetected in single image. Keywords – Deep Learning, CNN, FER-2013, IMDB, Keras, Haar feature-based cascade.AcknowledgementWe have immense pleasure in successful completion of this project on FacialExpression Recognition Using Deep Neural Networks. We would like to take this opportunity to express our gratitude to Dr. C P S Prakash, Principal of DSCE, for permitting us to utilize all the necessary facilities of the institution. We are also very grateful to our respected Vice Principal, HOD of Computer Science& Engineering, DSCE, Bangalore, Dr. Ramesh Babu D R, for his support andencouragement. We are immensely grateful to our respected and learned guide, Dr. Shubha Bhat, Associate Professor CSE, DSCE for their valuable help and guidance. We are extremelythankful to them for all the encouragement and guidance they have given us during everystage of the project. We would like to thank our project coordinators Dr. Vindhya M, Associate Professor, CSE, DSCE for their guidance and support. We are also thankful to all the other faculty and staff members of our department fortheir kind co-operation and help. Lastly, we would like to express our deep appreciation towards our classmates and ourfamily for providing us with constant moral support and encouragement. SHIVANI DUTT [1DS15CS099]MEENAKSHI BHAT [1DS15CS057]SHALINI [1DS15CS093]SHWETHA SANJAY SAVALGI [1DS15CS105]List of Figures4.1 Block Diagram of Architecture…………………………………………………….64.2 Simple and Deep Neural Network Diagram………………………..….………..74.4 Network Architecture……………………………………………………………… 84.5 Data Set Training………………….…………………………………………….….94.6 Flow Chart…………………………………………………………………………..105.2 Implementation details5.2.1 Organization of Implementation Files……….…………………………………126 Testing……………………………………….……………………………….…… 167 Results- Average Accuracy% for subject-independent………..…….………… 177.1 Average Accuracy% for subject-independent……………………….…………..187.2 Average Results for CK+ and MMI database……………………….…………..187.3 Training loss and classification accuracy on validation set.……….…………..197.4 Average Accuracy (%) on cross database………….…..………….………… 197.5 Better performance on MMI and FER2013………….………….………….……207.6 Predicted Label………….………….…………….…….………….………….……20Content1. Introduction…………………………………….……………………………….…12. Problem Statement……………………………………………………….………3. Literature Survey………………………………………………………….………33.1 Human FER from static image using shape and appearence………………..33.2 FER using Viola Jones algorithm and Fusion of PCA and ANN……………..33.3 Facial Expression Recognition…………………………………………..………43.4 Comparison of PCA and LDA Techniques………………….…………………..43.5 FER using Visual Saliency and Deep Learning……………………..………….54 . Architecture……………………………………………………….…………………64.1 Block Diagram………………………………………………………………..64.2 Deep Neural Architecture……………………………………………………74.3 Deep Neural Network Architecture…………………………………………74.4Network Architecture………………………………………………………….84.5Data Set Diagram……………………………………………………………..94.6Flow Chart…………………………………………………………………….105. Implementation…………………………………………………………..…………115.1 Hardware……………………………………………………………………..115.1.1Hardware………………………………………………………………115.1.2 Software………………………………………………………………115.2 Implementation Details……………………………………..………………125.2.1 Organization of Implementation Details………………………… 125.2.2 Installation……………………………………………………………135.2.3 Dataset Collection…………………………………………………..135.2.4 Loading FER Dataset……………………………………………….145.2.5 Training CNN model…………………………………………………145.2.6 Testing………………………………………………………………..156.Testing……………………………………………… ………………………………..167. Results…………………………………………………………………………..……178. Conclusion……………………………………………………………………………219. References……………………………………………………………………………2210.Appendix : Code…………………………………………………………………….