Temporal Segment Networks for Action Recognition in Videos

TL;DR

This paper introduces Temporal Segment Networks (TSN), a flexible framework for action recognition in videos that models long-range temporal structures, achieving state-of-the-art results on multiple benchmarks with efficient computation.

Contribution

The paper proposes a novel segment-based sampling and aggregation method within TSN, enabling effective learning of long-range temporal dependencies in videos for action recognition.

Findings

Achieved state-of-the-art accuracy on four benchmarks.

Efficiently models long-range temporal structures in videos.

Won the ActivityNet challenge 2016 video classification track.

Abstract

Deep convolutional networks have achieved great success for image recognition. However, for action recognition in videos, their advantage over traditional methods is not so evident. We present a general and flexible video-level framework for learning action models in videos. This method, called temporal segment network (TSN), aims to model long-range temporal structures with a new segment-based sampling and aggregation module. This unique design enables our TSN to efficiently learn action models by using the whole action videos. The learned models could be easily adapted for action recognition in both trimmed and untrimmed videos with simple average pooling and multi-scale temporal window integration, respectively. We also study a series of good practices for the instantiation of TSN framework given limited training samples. Our approach obtains the state-the-of-art performance on four challenging action recognition benchmarks: HMDB51 (71.0%), UCF101 (94.9%), THUMOS14 (80.1%), and ActivityNet v1.2 (89.6%). Using the proposed RGB difference for motion models, our method can still achieve competitive accuracy on UCF101 (91.0%) while running at 340 FPS. Furthermore, based on the temporal segment networks, we won the video classification track at the ActivityNet challenge 2016 among 24 teams, which demonstrates the effectiveness of TSN and the proposed good practices.