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

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

NeurIPS 2021 (Oral). Project page: https://facebookresearch.github.io/Motionformer

References in corpus (8)

• High-Speed Tracking with Kernelized Correlation Filters
• The Kinetics Human Action Video Dataset
• Linformer: Self-Attention with Linear Complexity
• Generating Long Sequences with Sparse Transformers
• Reformer: The Efficient Transformer
• Temporal Segment Networks: Towards Good Practices for Deep Action Recognition
• FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks
• TEA: Temporal Excitation and Aggregation for Action Recognition

Cited by in corpus (2)

• Robotic Vision for Human-Robot Interaction and Collaboration: A Survey and Systematic Review
• StarVQA: Space-Time Attention for Video Quality Assessment