Video Self-Stitching Graph Network for Temporal Action Localization

arXiv:2011.14598v4 Announce Type: replace
Abstract: 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 …

abstract arxiv challenge cs.cv datasets graph localization major network paper performance stitching temporal type variation video videos