Publication

Abstract :

Inspired by the new advancements in machine learning and search engine technology, it is possible to caption videos and index them for efficient managing in applications like YouTube, Dailymotion, Netflix, etc. This system can generate a set of tags which when put together will give a sensible translation of the video in text format. The key challenge is the semantic gap between identification of objects by the human brain and the computer AI. In this proposed model, we propose to transcribe the video clips based on the contents and construct a pipeline through the convolutional neural network (CNN) which will efficiently and accurately process the video and search for the contents. This would require processing of the video clip and indexing it in a database. The same indexing will enable the use of this technology in content-based image retrieval (CBIR) tasks. The user can search for videos based on the context, and the time spent is seeking through the video to search for the scene is solved. We use deep learning model to search through video content and show that it performs efficiently.