TSM: Temporal Shift Module for Efficient Video Understanding

TLDR

The explosive growth in video streaming creates a need for video understanding that balances high accuracy with low computational cost. This paper proposes a generic and effective Temporal Shift Module (TSM) that achieves both high efficiency and high performance. TSM shifts a subset of channels along the temporal axis to exchange information between neighboring frames, allowing it to be inserted into 2D CNNs for temporal modeling with zero extra computation or parameters, and it has been extended to an online version for real‑time low‑latency recognition and detection. TSM matches 3D CNN performance while retaining 2D CNN complexity, ranking first on the Something‑Something leaderboard and achieving 13 ms and 35 ms latency on Jetson Nano and Galaxy Note 8 for online recognition. Code is available at https://github.com/mit-han-lab/temporal-shift-module.

Abstract

The explosive growth in video streaming gives rise to challenges on performing video understanding at high accuracy and low computation cost. Conventional 2D CNNs are computationally cheap but cannot capture temporal relationships; 3D CNN based methods can achieve good performance but are computationally intensive, making it expensive to deploy. In this paper, we propose a generic and effective Temporal Shift Module (TSM) that enjoys both high efficiency and high performance. Specifically, it can achieve the performance of 3D CNN but maintain 2D CNN's complexity. TSM shifts part of the channels along the temporal dimension; thus facilitate information exchanged among neighboring frames. It can be inserted into 2D CNNs to achieve temporal modeling at zero computation and zero parameters. We also extended TSM to online setting, which enables real-time low-latency online video recognition and video object detection. TSM is accurate and efficient: it ranks the first place on the Something-Something leaderboard upon publication; on Jetson Nano and Galaxy Note8, it achieves a low latency of 13ms and 35ms for online video recognition. The code is available at: https://github. com/mit-han-lab/temporal-shift-module.