SyncAI.news, a Varaisys broadcasting
Groupwise Selective State-Space Filtering for Accurate and Streaming Action Boundary Detection
MB

Mustafa Bora \c{C}elik

· 1 min read

ResearcharXiv cs.CV

Groupwise Selective State-Space Filtering for Accurate and Streaming Action Boundary Detection

arXiv:2609.33400v1 Announce Type: new Abstract: Action boundary detection partitions untrimmed video into intervals without assigning action classes. We present a boundary-detection adapter operating on pre-extracted video features, learning temporal representations via groupwise selective scans. Learned group fusion and temporal modeling convert these into transition scores, which are decoded into boundary timestamps. Trained with boundary-time supervision, the class-agnostic model is evaluated on Breakfast, GTEA, and 50Salads using temporal tolerances and bipartite matching, achieving boundary $F_1$ scores of 0.457, 0.622, and 0.611. A stateful variant enables feature-streaming inference with zero neural look-ahead, one-sample peak confirmation, and bounded memory. Downstream systems can subsequently assign s

Original source

This story was published by arXiv cs.CV and written by Mustafa Bora \c{C}elik. SyncAI.news shows a preview; the complete article is on the publisher's site.

Read the full story on arxiv.org

Similar News