Video Data Retrieval TRIZ Cases
- Vehicle Video Query Triggers for Custom Event Detection
Natural-language queries are matched to LLM-generated video captions to detect relevant vehicle events without retraining static dashcam models.
- LLM-Guided Video Object Filtering for Accurate Text-Based Search
Text prompts are converted into filtering objects that guide video detectors, improving object identification without manual review.
- Digital Content Snapshot Search for Cross-App Time-Travel Retrieval
Captured screen snapshots are indexed with ML-extracted text to retrieve previously viewed cross-app content with fewer search steps and less overhead.
- Live Room Search Preview for Direct Content Navigation
Matched live-room recommendations and demonstrated-object details let users choose content before entering a room.
- Multi-tier Visual Search System with Ranked Object Model Segmentation
Segmenting object models by rank distributes recognition across tiers, resolving the contradiction between large database size and search time.
- Target-Video Co-Watched Clusters for Recommendation Diversity
Segmenting co-watched videos into keyword-based clusters improves recommendation diversity without increasing system complexity.
- Web Page Video Content Optimization via Analytics
A system optimizes web page content by analyzing user interactions with embedded video elements.
- Recall and Triggering System for On-Air Content Control
Embedding recall commands in video signals allows remote stations to insert local graphics, resolving the bottleneck of adding content after transmission.
- Automated Streaming Scheduling via RESTful APIs
Scheduling server manages live video streams using RESTful APIs to automate start and stop operations, eliminating manual intervention.
- Handheld Device Space Equipment Recognition
A handheld mobile wireless device scans space equipment images to identify items and personnel through a centralized database.