Long-Term Feature Banks for Detailed Video Understanding

Citations Over TimeTop 1% of 2019 papers

Abstract

To understand the world, we humans constantly need to relate the present to the past, and put events in context. In this paper, we enable existing video models to do the same. We propose a long-term feature bank—supportive information extracted over the entire span of a video—to augment state-of-the-art video models that otherwise would only view short clips of 2-5 seconds. Our experiments demonstrate that augmenting 3D convolutional networks with a long-term feature bank yields state-of-the-art results on three challenging video datasets: AVA, EPIC-Kitchens, and Charades. Code is available online.

Related Papers

• → EPIC AND PERFORMANCE - (F.) Macintosh, (J.) McConnell Performing Epic or Telling Tales. Pp. xii + 157, ills. Oxford: Oxford University Press, 2020. Cased, £58, US$77. ISBN: 978-0-19-884658-1. - (F.) Macintosh, (J.) McConnell, (S.) Harrison, (C.) Kenward (edd.) Epic Performances from the Middle Ages into the Twenty-First Century. Pp. xxii + 643, ills. Oxford: Oxford University Press, 2018. Cased, £120, US$160. ISBN: 978-0-19-880421-5.(2021)1 cited
• → Comments on “Epic and History”(2009)2 cited
• → Virgil and the Roman Epic - R. M. Henry: Virgil and the Roman Epic. Pp. 24. Manchester: University Press, 1938. Paper, is. 6d.(1939)
• → SEEING IN EPIC - H. Lovatt The Epic Gaze. Vision, Gender and Narrative in Ancient Epic. Pp. x + 414. Cambridge: Cambridge University Press, 2013. Cased, £70, US$110. ISBN: 978-1-107-01611-8.(2015)
• → Musing on the Past: Historical Epic and Epic History at Rome(2020)