Suggestion-Based Interactive Video Digest Design by User-System Cooperative Evolution


This paper proposes a suggestion-based interactive evolution method for video summarization, which attempts to enhance users' creative thought process without disturbing user edit operation. Although solutions of video summarization are time-varying, a few solutions can be evaluated in parallel. Therefore, the proposed method optimizes summarized video in background asynchronously with user operation. The background optimization is modeled as bi-objective optimization of user preference estimated by user operations and solution novelty. In this, Pareto solutions are stored to an archive and solutions are selected to suggest to the user based on α-domination. Experimental results using an eye-tracking device have revealed that the proposed method enhances convergent thinking rather than divergent thinking.