Deep Learning based Face Recognition using Convolution Neural networks

The term face recognition is the most powerful biometric technique to identify an individual. In this study, we have performed the task with the help of Convolutional Neural Networks. CNNs based deep architecture efficiently extracted features to identify individuals from a given set of images. In this technique, VGG16, EfficientNet-B0 and ResNet-50 have employed for feature extractions. Classical Machine Learning classification methods such as Support Vector Machine(SVMs), Logistic Regression, Decision Tree, Random Forest and Gaussian Naïve Bayes have been utilized for recognition of classes. We have made our classification on five individual classes. The results we have obtained for this task is quite well. For different network architecture, we have got results on the selected classifiers. Using VGG16 architecture, obtained accuracy for SVM with linear kernel is higher than other classifiers i.e. 97.81%. But Logistic regression performance is also good with 95.18%. On the contrast, EfficientNet-B0 architecture provides significant accuracy in all the classifiers but SVM and Logistic regression have obtained better result 98.68% and 98.25% respectively. ResNet50 also has good features and that’s why it acquired effective result in all the classifiers but logistic regression and random forest obtained the better outcomes with 98.68%. From this result it is clear that the classifiers have captured effective features for recognition of individuals.

• Supervision