Video Recognition via Segment-Based Feature Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video recognition methods for medium and long videos suffer from ignoring semantic associations and differences between video segments due to the extraction of a fixed number of frames, leading to insufficient features and reduced accuracy, which cannot be compensated by designing the feature extraction network.
Innovation Solution
A video recognition method that divides a video into segments, extracts frames from each segment, performs feature recognition, fuses the initial semantic feature blocks using a dynamic movement strategy, and determines the video type based on the fused target semantic feature block, ensuring that semantic associations and differences are captured.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a fixed number of frames are extracted from medium and long videos, then the processing complexity is reduced, but the semantic associations and differences between video segments are ignored, leading to insufficient features and reduced recognition accuracy
Solution Approach 1:
The video is divided into multiple segments, and frames are selectively extracted from each segment based on semantic features rather than using a fixed extraction pattern. This segmentation approach preserves semantic associations within segments while reducing unnecessary complexity from processing all frames uniformly.
Solution Approach 2:
Different extraction strategies are applied to different video segments based on their local characteristics. The system identifies key segments with important semantic information and extracts more frames from these regions, while using fewer frames from less critical segments, thereby optimizing the balance between feature sufficiency and processing complexity.
2Measurement precision
If more video frames are extracted to capture sufficient features, then the recognition accuracy improves, but the processing time and computational resources increase
Solution Approach 1:
The system performs preliminary analysis on video segments to identify which segments contain important semantic information before full frame extraction. This preliminary action allows the system to focus computational resources on extracting frames only from segments that contribute most to recognition accuracy, reducing overall processing time.
Solution Approach 2:
Instead of extracting a fixed number of frames uniformly from the entire video, the system extracts frames selectively from specific segments based on their semantic importance. This partial action approach ensures sufficient features are captured from critical segments while avoiding unnecessary extraction from less important segments, optimizing the trade-off between accuracy and processing time.
Data Source
AI summary
Provided are a video recognition method and apparatus, an electronic device, a medium and a computer program product. The video recognition method is described below. A to-be-recognized video is divided into at least two video segments; video frames are extracted from the at least two video segments, and feature recognition is performed on the video frames to obtain initial semantic feature blocks of the at least two video segments; each of the initial semantic feature blocks is fused, and a fused target semantic feature block is obtained; and a type of the to-be-recognized video is determined according to the fused target semantic feature block.