A Video Summarization Method Using Temporal Interest Detection and Key Frame Prediction

In this paper, a Video Summarization Method using Temporal Interest Detection and Key Frame Prediction is proposed for supervised video summarization, where video summarization is formulated as a combination of sequence labeling and temporal interest detection problem. In our method, we firstly built a flexible universal network frame to simultaneously predicts frame-level importance scores and temporal interest segments, and then combine the two components with different weights to achieve a more detailed video summarization. Extensive experiments and analysis on two benchmark datasets prove the effectiveness of our method. Specifically, compared with other state-of-the-art methods, its performance is increased by at least 2.6% and 4.2% on TVSum and SumMe respectively.