Introduction
We present a comparative study of the performance of several generalized feature extraction techniques when applied to face recognition. Firstly, deep convolutional neural networks (CNN) that have been trained on ImageNet were appropriated for face recognition. Additionally, traditional feature extraction methods, namely the eigenface approach as well as the histogram of oriented gradients (HOG), were investigated.
In this paper, we also attempt to combine several feature extraction techniques to form a joint representation that can be used for recognition.
Our results show that a combination of these feature extraction methods provides the best accuracy in comparison to any single technique.
Learn more
Read more at 👉 link