ESUMM: Event SUMMarization on Scale-Free Networks

In this digital era, events based video summarization has become an emerging topic in the field of computer vision which finds its applications in video data management, highlights of a game, video surveillance, etc. In this work, we propose a novel network-based approach for event summarization. First, the scale-free network is mapped to a neural network, and then dynamics of a complex video are determined by Chiavlo maps of the network. Second, we declare the neurons as hubs that are exceedingly well linked up as opposed to a majority who receive very short connections. Eventually, these hubs assume as the key-figures of the outcomes; then hubs and all its associated neurons counted for the issue summaries. Experiments show that the proposed approach outperforms the previous existing approaches on Precision and F-measure. The computing cost shows that the proposed model meets the criteria for real-time applications.