Temporal Action Detection with Frequency Attention Mechanism

Due to the variability of video length and action duration, the temporal action detection task faces the problem of blurred action boundaries that are difficult to capture accurately. To alleviate this problem, this paper proposes a Frequency Attention Mechanism (FAM) that adaptively models the frequency dependencies between video signal channels, enabling the model to better understand the frequency variations in the video and to handle the complexity of different action durations, thus enhancing the sensitivity and discriminative power of the action boundaries, and still providing powerful action recognition even in long video sequences Capabilities. Through comprehensive experimental validation on a series of representative benchmark datasets (e.g. THUMOS14 and ActivityNet1.3), our approach demonstrates significant performance improvement.

I. Introduction

As one of the core technologies in the field of video under standing, the importance of Temporal Action Detection (TAD) is not only reflected in the accurate recognition and classification of actions in videos, but also in the close integration with other tasks. For example, in the field of target tracking, TAD can be used to identify and locate the target in the video to provide accurate initialization information for target tracking; in the field of behavior recognition, TAD can be used to extract behavioral features in the video to improve the accuracy and robustness of behavior recognition. With the booming development of deep learning technology, research in the field of TAD has made great progress. However, despite the many achievements, TAD still faces challenges in practical applications. For example, issues such as the duration diversity of actions and the ambiguity of boundaries make temporal action detection still a challenging research area.