Hybrid Target Detection for Known and Unknown Object Classes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing target detection algorithms struggle to accurately recognize both predefined and unknown classes in images, leading to challenges in ensuring comprehensive detection accuracy.
Innovation Solution
A hybrid approach combining closed-set and open-set detection models is employed, where closed-set detection models with higher accuracy are used to recognize predefined classes, and open-set detection models identify unknown classes, with the results being fused to enhance overall detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a target detection model is trained by using a large amount of training data to recognize images of more classes, then the number of recognizable classes is improved, but it is difficult to ensure that all target objects in the image can be recognized and a large number of training samples are needed
Solution Approach 1:
The patent segments the detection task into two distinct models: a closed-set detection model for known classes and an open-set detection model for unknown classes. This segmentation allows each model to be optimized for its specific function, with the closed-set model achieving high accuracy on predefined classes using limited training data, while the open-set model handles novel classes without requiring extensive training samples for each new class.
Solution Approach 2:
The open-set detection model serves as a universal component that can detect both known and unknown classes. By designing the open-set model to identify novel classes without requiring specific training data for each class, the system achieves versatility across diverse object categories while avoiding the need to collect and train on large numbers of samples for every possible class.
2Measurement precision
If only closed-set detection is used to ensure high detection accuracy for predefined classes, then detection accuracy is improved, but unknown classes cannot be detected
Solution Approach 1:
The detection system is segmented into two specialized models: the closed-set detection model focuses exclusively on achieving high accuracy for predefined classes through targeted training, while the open-set detection model specializes in identifying unknown and novel classes. This functional segmentation allows each model to excel at its specific task without compromise.
Solution Approach 2:
The open-set detection model acts as an intermediary that bridges the gap between closed-set detection and unknown class recognition. It receives the same input image and provides complementary detection results for classes not covered by the closed-set model, thereby extending the system's capabilities to unknown classes while maintaining the high accuracy of the closed-set model for known classes.
3Adaptability or versatility
If only open-set detection is used to ensure detection of all target objects including unknown classes, then completeness of detection is improved, but detection accuracy decreases
Solution Approach 1:
The system segments the detection functionality into two models with distinct optimization goals. The closed-set detection model is trained specifically to achieve high accuracy on predefined classes, while the open-set detection model is optimized for detecting novel and unknown classes. By segmenting these functions, the system avoids the trade-off where a single model must compromise between accuracy and completeness.
Solution Approach 2:
The patent merges the results from both the closed-set detection model and the open-set detection model to produce the final detection output. This combination allows the system to leverage the high accuracy of the closed-set model for known classes while simultaneously incorporating the comprehensive detection capabilities of the open-set model for unknown classes, achieving both accuracy and completeness in the final result.
Data Source
AI summary
This disclosure provides a target detection method. In the method, a closed-set detection is performed on an image via a closed-set detection model to obtain a first detection result. An open-set detection is performed on the image via an open-set detection model to obtain a second detection result. Accuracy of the first detection result is higher than accuracy of the second detection result. The first detection result and the second detection result are merged to obtain a target detection result of the image.