Automatic Index Point Detection and Semantic Indexing of Educational Videos

Abstract

The purpose of this study is to identify the need for automatic video indexing and present a system to automatically construct a digital video index for better information retrieval. This project aims to solve a specific problem i.e., extracting keywords from video frames so that a lecture may be split. An index point is a term that additionally carries the relative video time instance's timestamp. To extract the video's textual information, Google's OCR Engine is used. The research is confined to educational videos that include presentation slides in English only. We have tested three methods for index point detection and finally concluded that automatic index points detection and extraction using deep learning neural network outperforms the other two methods in terms of speed and accuracy.