Proceedings of the International Display Workshops Volume 30 (IDW '23)
Dec 6 - Dec 8, 2023TOKI MESSE Niigata Convention Center
Proceedings of the International Display Workshops Volume 30 (IDW '23)
Dec 6 - Dec 8, 2023TOKI MESSE Niigata Convention Center
[MVS3/INP4-3 (Invited)]Frame-Wise Action Recognition for Skeleton-Based Anomaly Detection
*Hiroaki Tani1, Tomoyuki Shibata1(1. Toshiba Corporation (Japan))
Keywords:
action recognition,computer vision,deep learning,anomaly detection
We propose a novel method for recognizing human actions from a single frame. Our approach utilizes the “action-ness” obtained by a pretrained model, enabling effective training for frame-wise action recognition without additional data. We also introduce several anomaly detection applications utilizing our method.