AI Image Retrieval Using Deep Supervised Hash Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image retrieval methods struggle in complex scenarios due to the inability to distinguish easily confused entities, failing to accurately capture image similarity relationships, and treating all samples equally without providing adequate attention to complex samples.
Innovation Solution
A method and system utilizing artificial intelligence and a trained convolutional neural network to extract image features, employing hashing methods to differentiate confusable entities and optimize similarity relationships, while providing attention to complex samples through a deep supervised hash learning approach.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If popular article retrieval methods treat all samples equally, then the method is simple to implement, but the generalization performance deteriorates in complex scenarios
Solution Approach 1:
The patent applies local quality by differentiating sample processing based on complexity. Simple samples use standard feature extraction, while complex samples undergo additional processing including confusion entity identification and similarity relationship mining. This localized adaptation resolves the contradiction by maintaining simplicity for easy cases while enhancing performance for complex cases.
Solution Approach 2:
The system dynamically adjusts the retrieval process based on sample complexity assessment. It transitions from a static uniform approach to a dynamic multi-path processing system that adapts its complexity level according to the input sample characteristics, thereby improving generalization performance without uniformly increasing complexity across all cases.
2Productivity
If popular article retrieval methods use standard feature representation, then the processing is efficient, but the ability to distinguish confusable entities deteriorates
Solution Approach 1:
The patent segments the feature extraction process into multiple stages: initial feature extraction using standard methods, followed by confusion entity identification, and then refined feature generation. This segmentation allows the system to maintain efficiency in the initial stage while improving distinction accuracy in subsequent stages only where needed.
Solution Approach 2:
The patent introduces an intermediary component (confusion entity identification module) that detects when standard feature representation is insufficient. This intermediary triggers additional processing only for cases where it is needed, thereby maintaining overall efficiency while improving entity distinction accuracy for difficult cases.
3Device complexity
If existing article retrieval methods do not mine similarity relationships, then the computational complexity is low, but the accuracy of image similarity deteriorates
Solution Approach 1:
The patent applies partial action by selectively mining similarity relationships only for complex samples identified through the confusion entity detection mechanism. Instead of applying similarity mining to all samples (excessive action), it applies the computation only where needed (partial action), thereby balancing computational complexity with similarity accuracy.
4Measurement precision
If article retrieval in complex scenarios requires more accurate image similarity, then the retrieval accuracy improves, but the computational cost increases
Solution Approach 1:
The patent performs preliminary assessment of sample complexity and pre-identifies confusion entities before conducting full similarity mining. This preliminary action allows the system to prepare only for cases that will require intensive processing, thereby reducing overall computational cost while maintaining high retrieval accuracy for complex samples.
Data Source
AI summary
A method and system for rapid retrieval of target images based on artificial intelligence, obtaining a template image and a plurality of known labels corresponding to the template image; extracting an image to be detected from a target image database; inputting both the image to be detected and the template image into a trained convolutional neural network, and outputting a hash code of the image to be detected and a hash code of the template image; obtaining a similarity between the images based on a Hamming distance between the hash codes, then selecting one or more images to be detected with the similarity higher than a set threshold as a retrieval result to output. Accordingly, the method and system is able to better cope with the retrieval of items in complex scenarios.