Video Data Querying TRIZ Cases
- AOT Tuning Data Classification Using AV Operating State Correlation
AV operating state data and forward-backward record matching reduce AOT tuning data, improving media ratings accuracy.
- Video Feature Fusion Across Receptive Fields for Accurate Querying
Multiple receptive fields and cross-attention merge local and global video features to improve classification and video query accuracy.
- Video Moment Retrieval With Cross-Attention Query Weighting
Cross-attention weighting links text queries to video features, improving retrieval of relevant moments and highlight segments.
- Security Video Search Using Natural Language and Shared Embeddings
Maps sampled video frames and natural language queries into a shared embedding space to speed event search across large security video sets.
- Voice-Controlled Display Content Name Mapping for Accurate Commands
Maps abbreviations and varied spoken content names to official titles so voice-controlled displays return the user-intended result.
- Video Search Term Presentation for Comment-Relevant Discovery
Separating video-page and comment-page search terms keeps recommendations tied to viewed content while capturing extended search intent.