Less is More: Learning Highlight Detection from Video Duration
This addresses the need for scalable highlight detection in video browsing, though it is incremental as it builds on existing unsupervised approaches.
The paper tackles the problem of expensive supervision in video highlight detection by proposing an unsupervised method that uses video duration as an implicit signal, training on 10M Instagram videos and achieving substantial improvements on public benchmarks.
Highlight detection has the potential to significantly ease video browsing, but existing methods often suffer from expensive supervision requirements, where human viewers must manually identify highlights in training videos. We propose a scalable unsupervised solution that exploits video duration as an implicit supervision signal. Our key insight is that video segments from shorter user-generated videos are more likely to be highlights than those from longer videos, since users tend to be more selective about the content when capturing shorter videos. Leveraging this insight, we introduce a novel ranking framework that prefers segments from shorter videos, while properly accounting for the inherent noise in the (unlabeled) training data. We use it to train a highlight detector with 10M hashtagged Instagram videos. In experiments on two challenging public video highlight detection benchmarks, our method substantially improves the state-of-the-art for unsupervised highlight detection.