Spatio-temporal human action localization in indoor surveillances

Zihao Liu, Danfeng Yan, Yuanqiang Cai, Yan Song

Abstract: Highlights•Developed a novel benchmark for real-world indoor surveillance, with human-centric, frame-level annotations.•Our benchmark suits security well due to diverse scenes and complex actions.•Evaluating state-of-the-art methods on our dataset to report performance.•Conducting performance analysis on HIA exposes challenges in spatio-temporal localization.