A ResNet deep learning based facial recognition design for future multimedia applications. (December 2022)

• Record Type: Journal Article
• Authors: Durga, B Kanaka; Rajesh, V.
• Is Part Of: Computers & electrical engineering. Volume 104:Part A(2022)
• Publication Date: 2022-12
• Subjects: Micro-facial emotions -- Maskable features extraction -- 2D-ResNet CNN -- Deep learning; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.108384
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: A real time facial emotions recognition application using deep learning technology. Facial emotions detection even facial images having face mask and spectacles. Facial expression features extraction during training process with ResNet technology. Deep learning based continues miniaturization of facial emotions and tracking. Classification and feature extraction mechanism can help the Facial emotions. Abstract: CNN deep-learning models are the most trusted computational approaches in industrial, medical, and intelligent applications. In this research work, a deep micro-facial emotion has been recognised to employ CNN approach. Multi-class feature extraction has been initiated by the 2D-ResNet convolutional neural network. In order to detect the maskable images of facial emotions, a 2D-ResNet CNN multi-class classifier has been proposed. This novel model has reduced overfitting and controversial problems, which are verified on the Python 3.7.0 software tool. The public dataset JAFFE (open source) has been used to train the 2D-ResNet deep learning model. The flask, mysql, fastai, and jsonify python packages are compatible and feasible with the 2D-ResNet Deep Learning architecture. The proposed method practices better performance metrics in terms of accuracy of 99.3%, recall of 99.12%, F1 score of 0.98% and sensitivity of 99.16%, which are figured out from the confusion matrix, where only 5 failures have been observed when tested with 1 lakh images.