Quality-Aware Recognition Model Selection for Video Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video processing systems fail to maintain recognition accuracy when video quality fluctuates due to compression or transmission over networks, leading to erroneous recognition.
Innovation Solution
A video processing system that includes multiple recognition models trained for different video quality parameters, allowing selection of the most suitable model based on the current video quality parameters for improved recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single recognition model is used for all video qualities, then the system complexity is low, but the recognition accuracy decreases when video quality fluctuates due to compression or transmission
Solution Approach 1:
The patent divides the recognition system into multiple specialized models, each trained for specific video quality ranges. Instead of using one general model, the system segments the video quality spectrum into different segments (high quality, medium quality, low quality) and creates dedicated recognition models for each segment, allowing optimal performance across varying quality conditions
Solution Approach 2:
The patent implements dynamic model selection where the system adapts its behavior based on input video quality. The quality assessment unit continuously evaluates video quality parameters and dynamically switches between different recognition models according to the current quality level, enabling the system to maintain high accuracy across fluctuating quality conditions
2Reliability
If multiple recognition models for different video qualities are implemented, then the recognition accuracy improves, but the device complexity increases
Solution Approach 1:
The patent performs preliminary training of multiple recognition models with different quality parameters before actual operation. Each model is pre-trained on video data corresponding to specific quality ranges, so that during runtime, the system simply needs to select the pre-trained model that matches the current video quality, avoiding the need for real-time model training or complex adaptive algorithms
3Productivity
If video is transmitted via network with compression, then the transmission efficiency improves, but the video quality deteriorates leading to erroneous recognition
Solution Approach 1:
The patent changes the parameter of video quality expectation based on transmission conditions. Instead of assuming high video quality, the system adjusts its expectations and selects recognition models appropriate for compressed, lower-quality video. This parameter adaptation allows the system to maintain recognition accuracy even when network compression degrades video quality
Data Source
AI summary
A video processing system (10) according to the present disclosure includes a recognition model (M1), a recognition model (M2), a recognition model (M3), and a recognition model (M4) that have learned video learning data corresponding to different video quality parameters, for each of the video quality parameters; and a selection unit (11) that selects a recognition model that performs recognition regarding a target included in the video input data to be input, from among the recognition model (M1), the recognition model (M2), the recognition model (M3), and the recognition model (M4), according to a video quality parameter of the video input data.