Face Detection for Real World Application

Abstract

Face Detection has become a very prevalent issue in Machine Learning, not only in machine learning but in any field, one can think of. Due to this, it has gained a wide fan base and many people are working every day to improve the accuracy of object detection models using deep learning. But this improved performance comes at the price of increased computational overhead, which limits the ability of a machine learning model to be utilized on devices having small Graphical Processing Units. The core intent of this paper is to compare computation time for models such as Histogram of Oriented gradients (0.4 seconds) and ResNet (48.5 seconds) with BlazeFace (0.09 seconds), a model developed by google in the year 2020 and is a mobile device friendly model and fits well with real time application which need instant feedback and on top of that cannot handle bulky computations required for deep learning models