Abstract:
Object Detection is a fundamental problem in Computer Vision. Face Detection is an important and intriguing type of object detection that is being extensively used in day-to-day activities like in video surveillance, online social media, and robotics. Researchers have made tremendous progress in this eld by developing state-of-the-art detectors that, over the years, have increased in accuracy and speed of detection. But these detectors are only suited for some conditions. For example: detecting frontal faces, faces that are not occluded, well-lit faces, etc. They also don't work for faces in images of other spectra like near-infrared. Moreover, detectors that perform well in terms of accuracy, lack of speed and vice-versa. This research project aims to assess the strengths and weaknesses of existing state-of-the-art detectors, build a detector that works for multiple unconstrained conditions like occlusion, illumination, crowd, etc., and nd a suitable tradeoff between accuracy and speed of detection.
