Supervised Data Generation for AI Image Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating supervised data for AI image recognition systems are inefficient as they require manual preparation of weakly supervised data for training, which limits the efficiency of the data generation process.
Innovation Solution
A supervised data generation system and method that includes a shape information acquisition unit, an image generation condition setting unit, an image generation unit, and a supervised data output unit, which generates simulation images with position information as annotation, allowing for automated and efficient supervised data creation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation is used to create supervised data, then annotation accuracy can be ensured, but the data generation efficiency is low
Solution Approach 1:
The patent uses simulation images as copies of real object shapes to generate supervised data. By copying the essential geometric features of objects into simulated environments, the system produces annotated data without manual intervention while maintaining structural accuracy
Solution Approach 2:
The patent replaces the manual mechanical process of annotation with an automated computer-based simulation system. The image generation unit automatically creates simulation images with embedded position information, substituting human labor with computational processes
2Productivity
If AI is used to add annotation information, then data generation efficiency is improved, but the system complexity increases
Solution Approach 1:
The patent extracts only the essential position information from complex image data generation processes. By focusing solely on generating position annotations rather than full image processing, the system reduces complexity while maintaining efficiency
Solution Approach 2:
The patent introduces simulation images as an intermediary between real objects and annotation data. This intermediary layer simplifies the process by providing a controlled environment where position information can be automatically generated without dealing with the complexity of real-world image variability
Data Source
AI summary
The supervised data generation system according to the present disclosure includes a shape information acquisition unit that acquires shape information recording the shape of a target object, an image generation condition setting unit that sets image generation conditions for a simulation image of the target object, and a shape information and an image generation unit that generates a simulation image of a target object and position information of the target object in the simulation image based on image generation conditions; and an image generator that adds position information to the simulation image as annotation information; and a supervised data output unit that outputs a simulation image.