Real-time face detection using histogram of oriented gradients, convolutional neural networks and feature extraction to enhance the online shopping experience

Abstract

Recent trends in technology shows that computer vision has become an integral part of all artificial intelligence projects. Ranging from a normal barcode scanner to real-time CCTV surveillance and to self-driving cars, the applications of computer vision has become numerous in today’s world. With the development of deep learning technologies and AI, computer vision has grown from simple image detection from saved images to real-time recognition of objects from raw camera feed. The introduction of the Histogram of Oriented Gradient (HOG) method improved the process of computer vision to a great magnitude. This research aims to use the HOG method for image detection in real-time and to compare the efficiency of the HOG - Linear Support Vector Machines model with the Convolutional Neural Network algorithm, which is now the de facto standard in real-time detection. The research further intends to extract features from the detected image using the ensemble of regression trees method and then use image processing on it to enhance customer experience while shopping online. This research can thus be developed into a real-world application that can find its place in commercial markets, websites/mobile applications and online shopping platforms.