Smart Face Detection and Recognition in Illumination Invariant Images using AlexNet CNN Compare Accuracy with SVM

Aim: In many recent technologies it shows that deep learning based face recognition gets best results even surpassing human capability however many models suffer from illumination images. The aim of this study is to recognize faces in different illumination invariant images using deep learning and machine learning techniques. Materials and methods: 20 sample images of illumination invariant image dataset, totally 40 samples trained using deep learning method Alexnet Convolutional Neural Network (ACNN) and Machine learning method Support vector machine (SVM) with G Power 80% and alpha 0.05. Results: ACNN has achieved 98.4 percent accuracy when compared to SVM which has attained 86 percent in face recognition. Attained significance value for two tailed is 0.0001 (p< 0.05). Conclusion: For the given set of images, ACNN performs better compared to SVM.