A Conceptual Architecture for Deep-learning-based Video-Objects Retrieving System

Dinh-Lam Pham, Viet-Vu Vu, Byeongnam Yoon, Kyoungsook Kim, Kwanghoon Pio Kim

Abstract: Finding content in videos or video-objects retrieving systems has been a matter of concern for many years until now. Regularly, to retrieve and explore content in videos, the traditional solution is that we oblige to use the human eyes and examine those videos manually until we obtain the content that appears in the video. To speed up the search, we can fast-forward the video or predict when the content appeared. Nevertheless, this method often consumes a lot of time, especially when we need to retrieve content in a large number of Videos. In this paper, we introduce a deep-learning-based architecture that allows retrieving objects in videos instantly to search objects in CCTV cameras. This solution promises to bring high efficiency in finding video content, thereby saving costs and human resources to operate related systems.