Long-Term Feature Banks for Detailed Video Understanding

by   Chao-Yuan Wu, et al.

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.



page 1

page 2

page 6


ECO: Efficient Convolutional Network for Online Video Understanding

The state of the art in video understanding suffers from two problems: (...

Towards Long-Form Video Understanding

Our world offers a never-ending stream of visual stimuli, yet today's vi...

Taylor saves for later: disentanglement for video prediction using Taylor representation

Video prediction is a challenging task with wide application prospects i...

MeMViT: Memory-Augmented Multiscale Vision Transformer for Efficient Long-Term Video Recognition

While today's video recognition systems parse snapshots or short clips a...

Sequence Graph Transform (SGT): A Feature Extraction Function for Sequence Data Mining (Extended Version)

The ubiquitous presence of sequence data across fields such as the web, ...

Super-reflective Data: Speculative Imaginings of a World Where Data Works for People

It's the year 2020, and every space and place on- and off-line has been ...

Beyond Goldfish Memory: Long-Term Open-Domain Conversation

Despite recent improvements in open-domain dialogue models, state of the...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.