Object-to-Text Conversion for Precise Positional Relation Description
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems struggle to accurately describe the positional relations between objects in drawings, leading to variations in explaining these relations and difficulties in determining overlap states between objects, especially when using graphic drawing programs.
Innovation Solution
A method and system that convert objects into text by extracting coordinate data, determining positional relations, converting these into graph data, and generating text using an image processing unit, feature extraction unit, graph generation unit, and text generation unit, with a GUI for display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If image recognition is used to describe positional relations between objects, then automation is improved, but measurement precision deteriorates due to variations in explaining positional relations
Solution Approach 1:
The patent introduces graph data as an intermediary representation between image recognition and natural language generation. The graph data structure with vertices and edges provides a standardized, precise way to represent positional relations, serving as a mediator that bridges the gap between automated image analysis and accurate textual description.
Solution Approach 2:
The patent transforms positional relation information from continuous spatial coordinates into discrete graph data parameters (vertices, edges, and their attributes). This parameter transformation enables precise representation of positional relations by converting vague spatial concepts into structured, computable data elements with defined properties.
2Ease of operation
If graphic drawing programs are used to display objects, then ease of operation is improved, but measurement precision deteriorates due to overlap ambiguities
Solution Approach 1:
The patent uses graph data as an intermediary to resolve overlap ambiguities in graphic drawings. By representing objects and their spatial relationships in a structured graph format, the system can precisely determine overlap states without relying on visual interpretation of potentially ambiguous graphical displays.
Solution Approach 2:
The patent segments the complex task of determining overlap states into discrete graph data processing steps. By breaking down the spatial relationship analysis into vertex and edge operations on graph structures, the system can systematically and precisely determine overlap states without the ambiguities inherent in visual graphic interpretation.
Data Source
AI summary
Text is generated from an object. Text is generated from a first object. The first object includes a second object and a third object. A step of detecting coordinate data of the second object is included. A step of detecting coordinate data of the third object is included. A step of extracting positional relation between the second object and the third object from coordinate data is included. A step of converting the extracted positional relation into graph data is included. A step of generating text about the positional relation between the second object and the third object from graph data is included.