The CERTH-ITI paper “Unsupervised Video Summarization via Attention-Driven Adversarial Learning”, by E. Apostolidis, E. Adamantidou, A. Metsai, V. Mezaris, I. Patras, presented at the 26th Int. Conf. on Multimedia Modeling (MMM2020), Daejeon, Korea, received the MMM2020 Best Paper Award.

The awarded paper presents a new video summarization approach that integrates an attention mechanism to identify the significant parts of the video, and is trained unsupervisingly via generative adversarial learning. This makes it possible to train the proposed deep learning architecture without requiring manually-generated video summaries as training data; and, the proposed combination of attention and generative adversarial learning enables the creation of more representative and engaging video summaries. This work is applicable in sectors such as broadcasting, video archives, web-based video delivery and search, and others.