Multi-Model Feature Embedding for Accurate Similar Image Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image search techniques for electronic commerce platforms lack accuracy and speed in finding similar images based on product images, particularly when dealing with complex data.
Innovation Solution
An information processing apparatus that generates a compounded feature vector by concatenating feature vectors from multiple learning models, including first and second feature predictive models, a gender predictive model, and a color predictive model, to enhance the accuracy of similar image search.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single learning model is used to generate feature vectors for image search, then the system complexity is low, but the search accuracy is insufficient
Solution Approach 1:
The patent combines multiple learning models (first feature predictive model, second feature predictive model, attribute predictive model, color predictive model) into a unified feature extraction system. Each model contributes specific feature vectors that are concatenated to form a comprehensive compounded feature vector, thereby improving search accuracy while managing system complexity through modular integration.
Solution Approach 2:
The patent creates a composite feature representation by concatenating feature vectors from different learning models (classification features, attribute features, color features) into a single compounded feature vector. This composite approach integrates diverse feature types to achieve higher search accuracy, analogous to using composite materials to combine properties of different materials.
2Measurement precision
If multiple learning models are used to generate comprehensive feature vectors, then the search accuracy improves, but the data processing complexity increases
Solution Approach 1:
The patent segments the feature extraction process into multiple specialized learning models, each responsible for extracting specific types of features (classification, attributes, color). This segmentation allows each model to focus on particular aspects of image analysis, improving overall feature extraction accuracy while making the complex processing task more manageable through division of labor.
Solution Approach 2:
The patent introduces a concatenation unit as an intermediary component that integrates feature vectors from multiple learning models into a unified compounded feature vector. This intermediary structure simplifies the complexity of handling multiple separate feature sets by providing a standardized integration mechanism that combines diverse features into a single coherent representation.
3Measurement precision
If comprehensive feature vectors from multiple models are used, then the similar image search accuracy improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary feature extraction by generating feature vectors from multiple learning models in advance, before the actual similarity search. The compounded feature vectors are pre-computed and stored, allowing the search process to efficiently compare pre-prepared representations rather than processing raw images during search operations, thereby reducing real-time processing time while maintaining high accuracy.
Data Source
AI summary
An information processing apparatus (1) includes: an acquisition unit (101) configured to acquire an object image that contains a target object; a generation unit (102;103;104;105) configured to generate a plurality of feature vectors for the object by applying the object image to a plurality of learning models; a concatenation unit (106) configured to concatenate and embed the plurality of feature vectors into a common feature space to generate a compounded feature vector in the feature space; and a search unit (107) configured to search for a similar image that is similar to the object image, using the compounded feature vector.