Video Self-Stitching Graph Network for Temporal Action Localization
This work provides an incremental improvement in temporal action localization for the computer vision community, specifically by enhancing the detection of short actions.
The paper addresses the challenge of localizing short actions in videos, which typically have the lowest performance despite being prevalent. They propose a Video Self-Stitching Graph Network (VSGN) that magnifies short video segments and stitches them with original clips, achieving state-of-the-art overall performance on THUMOS-14 and ActivityNet-v1.3 datasets while significantly improving short action localization.
Temporal action localization (TAL) in videos is a challenging task, especially due to the large variation in action temporal scales. Short actions usually occupy a major proportion in the datasets, but tend to have the lowest performance. In this paper, we confront the challenge of short actions and propose a multi-level cross-scale solution dubbed as video self-stitching graph network (VSGN). We have two key components in VSGN: video self-stitching (VSS) and cross-scale graph pyramid network (xGPN). In VSS, we focus on a short period of a video and magnify it along the temporal dimension to obtain a larger scale. We stitch the original clip and its magnified counterpart in one input sequence to take advantage of the complementary properties of both scales. The xGPN component further exploits the cross-scale correlations by a pyramid of cross-scale graph networks, each containing a hybrid module to aggregate features from across scales as well as within the same scale. Our VSGN not only enhances the feature representations, but also generates more positive anchors for short actions and more short training samples. Experiments demonstrate that VSGN obviously improves the localization performance of short actions as well as achieving the state-of-the-art overall performance on THUMOS-14 and ActivityNet-v1.3.