Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers

TL;DR

This paper introduces trajectory attention for video transformers, modeling motion paths to improve dynamic scene understanding, and proposes a method to reduce computational complexity, achieving state-of-the-art results in video action recognition.

Contribution

It presents a novel trajectory attention mechanism and a computational efficiency method for video transformers, enhancing dynamic scene modeling and performance.

Findings

Achieved state-of-the-art results on Kinetics, Something--Something V2, and Epic-Kitchens datasets.

Demonstrated the effectiveness of trajectory attention in modeling motion paths.

Reduced computational and memory requirements for high-resolution or long videos.

Abstract

In video transformers, the time dimension is often treated in the same way as the two spatial dimensions. However, in a scene where objects or the camera may move, a physical point imaged at one location in frame $t$ may be entirely unrelated to what is found at that location in frame $t+k$. These temporal correspondences should be modeled to facilitate learning about dynamic scenes. To this end, we propose a new drop-in block for video transformers -- trajectory attention -- that aggregates information along implicitly determined motion paths. We additionally propose a new method to address the quadratic dependence of computation and memory on the input size, which is particularly important for high resolution or long videos. While these ideas are useful in a range of settings, we apply them to the specific task of video action recognition with a transformer model and obtain state-of-the-art results on the Kinetics, Something--Something V2, and Epic-Kitchens datasets. Code and models are available at: https://github.com/facebookresearch/Motionformer