基于事件一致性的视听事件检测方法及装置
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2022-11-25
- Publication Date
- 2026-07-17
AI Technical Summary
Existing audiovisual event detection methods have shortcomings in background fragment recognition and event semantic consistency, resulting in poor performance and ignoring the semantic consistency of events in the same video.
This paper proposes an audiovisual event detection method based on event consistency. It adopts audiovisual joint learning and semantic consistency modeling, including feature encoding at the segment level and semantic guidance at the video level. Semantic consistency modeling is performed using a cross-modal event representation extractor and GRU. It combines fully supervised and weakly supervised background class screening loss functions and inter-segment smoothing loss.
It improves the accuracy and discriminative power of audiovisual event detection, outperforming existing methods, especially in fully and weakly supervised tasks on the AVE dataset.
Smart Images
Figure CN115861879B_ABST