Image Tagging via Relevance Analysis and Combined Tags
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image tagging methods struggle to effectively associate and utilize image tags related to objects and their attributes or motions, leading to inefficient search results due to the lack of interconnectivity and relevance analysis between generated tags.
Innovation Solution
An image tag generating model is used to obtain multiple tags, followed by determining the relevance between them through area distribution maps and masks or bounding boxes, and generating a combined tag with interconnected tags based on relevance thresholds and search query probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple image tags are generated independently without relevance analysis, then the quantity of tags increases, but the relevance and accuracy of tag associations deteriorates
Solution Approach 1:
The patent implements feedback by analyzing the relevance between generated image tags and using this analysis to refine and optimize tag combinations. The system generates initial tags, evaluates their relevance through area distribution map comparisons, and uses this feedback to create optimized combined tags that improve overall tagging accuracy and search effectiveness.
Solution Approach 2:
The patent introduces an intermediary mechanism in the form of area distribution maps and relevance analysis modules that mediate between independently generated tags. These intermediaries evaluate the spatial and semantic relationships between tags, determining which tags should be combined and how they should be associated to maintain high relevance while increasing tag quantity.
2Ease of manufacture
If image tags are generated without analyzing spatial relationships, then the generation process is simpler, but the accuracy of tag-object associations deteriorates
Solution Approach 1:
The patent adds a spatial dimension to tag generation by introducing area distribution maps that represent the spatial relationships between objects and tags. Instead of simply generating tags based on object detection, the system creates 2D spatial representations showing where each tag applies within the image, thereby improving association accuracy without significantly complicating the generation process.
3Adaptability or versatility
If all possible tag combinations are created, then the completeness of tag coverage improves, but the complexity of tag processing increases
Solution Approach 1:
The patent applies partial action by creating only the necessary tag combinations rather than all possible combinations. The relevance analysis module identifies and processes only those tag pairs or groups that have meaningful spatial and semantic relationships, eliminating redundant processing while maintaining complete coverage of important tag associations.
4Productivity
If image tags lack interconnectivity, then the tagging process is faster, but the effectiveness of search retrieval deteriorates
Solution Approach 1:
The patent implements preliminary action by pre-establishing the interconnectivity structure through relevance analysis and area distribution map comparisons before the actual search retrieval process. The system pre-processes and stores the relationships between tags, so that during search operations, the pre-established connections can be quickly utilized without adding processing delays.
Data Source
AI summary
A method of performing image tagging may include obtaining a plurality of image tags from an image using an image tag generating model, determining a degree of relevance between the plurality of image tags based on a plurality of area distribution maps respectively corresponding to each image tag of the plurality of image tags, generating a combined tag with interconnected image tags therein based on the degree of relevance, and performing image tagging on the image using the plurality of image tags and the combined tag.