Video Digest Scene Extraction Using Shifted Importance Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for automatically extracting important scenes in video digests using machine learning struggle to include related scenes close to the important scenes, leading to incomplete or unsatisfactory video digests.
Innovation Solution
A video processing device and method that utilizes multiple models trained with shifted correct answer data to calculate and integrate importance scores, allowing for the extraction of important scenes and their related scenes by adding importance values at different time intervals, generating a more comprehensive digest video.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single machine learning model is used to extract important scenes, then the extraction process is simple and fast, but related scenes close to the important scenes are not appropriately included in the digest video
Solution Approach 1:
The patent divides the scene extraction task into multiple independent models, each responsible for detecting important scenes from different perspectives or temporal contexts. By segmenting the detection function across multiple models and integrating their results, the system comprehensively identifies both important scenes and related scenes without sacrificing extraction efficiency
Solution Approach 2:
The patent combines the detection results from multiple machine learning models through integration processing. By merging the outputs of individual models that detect important scenes from different angles, the system achieves a more complete identification of related scenes while maintaining the efficiency benefits of modular model execution
2Loss of information
If multiple models are used to calculate importance, then related scenes are included in the digest video, but the processing complexity and computational load increase
Solution Approach 1:
The patent segments the complex detection task into multiple specialized models, each handling a specific aspect of scene importance detection. This segmentation reduces the complexity burden on individual models while collectively achieving comprehensive scene identification through their integrated results
Solution Approach 2:
The patent designs multiple models with universal applicability to different video content types and scenarios. Each model is trained to detect important scenes through slightly different approaches, and their combined results provide a universal solution for comprehensive scene extraction across various video genres and contexts
Data Source
AI summary
In a video processing device, a video acquisition means acquires a material video. An importance calculation means calculates importance in the material video using a plurality of models. An importance integration means integrates the importance calculated using the plurality of models. A generation means extracts important scenes in the material video based on the integrated importance and generates a digest video including the extracted important scenes.