Spott

Abstract

Spott is an innovative second screen mobile multimedia application which offers viewers relevant information on objects (e.g., clothing, furniture, food) they see and like on their television screens. The application enables interaction between TV audiences and brands, so producers and advertisers can offer potential consumers tailored promotions, e-shop items, and/or free samples. In line with the current views on innovation management, the technological excellence of the Spott application is coupled with iterative user involvement throughout the entire development process. This article discusses both of these aspects and how they impact each other. First, we focus on the technological building blocks that facilitate the (semi-) automatic interactive tagging process of objects in the video streams. The majority of these building blocks extensively make use of novel and state-of-the-art deep learning concepts and methodologies. We show how these deep learning based video analysis techniques facilitate video summarization, semantic keyframe clustering, and (similar) object retrieval. Secondly, we provide insights in user tests that have been performed to evaluate and optimize the application’s user experience. The lessons learned from these open field tests have already been an essential input in the technology development and will further shape the future modifications to the Spott application.