Texture Mapping 3D Models for Synthetic Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods require extensive training data and time to accurately infer the shape, center position, or type of objects with complicated shapes or patterns, making them inefficient for real-world applications.
Innovation Solution
A training data generation device that acquires partial image information, 3D model information, and rendering conditions to generate two-dimensional image information by texture-mapping and rendering 3D models, reducing the training time required for generating a trained model capable of accurately inferring object features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If CG-generated training data with simple 3D models is used, then data generation efficiency is improved, but inference accuracy for objects with complicated shapes or patterns deteriorates
Solution Approach 1:
The patent uses texture mapping to copy real photographed images onto CG 3D models. The texture mapping unit maps textures from photographed images to corresponding regions on CG models, transferring complex patterns and details from real objects to virtual models. This allows the system to maintain simple CG geometries while incorporating realistic surface details through texture copying.
Solution Approach 2:
The patent creates composite training data by combining CG-generated image information with photographed image textures. The system synthesizes training images that contain both the structured geometry of CG models and the complex visual patterns of real objects through texture mapping, effectively creating a composite that leverages the advantages of both approaches.
2Measurement precision
If more training data is collected to improve inference accuracy for complicated objects, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The system copies textures from a limited set of photographed images and applies them to multiple CG models with varying geometries, lighting conditions, and camera angles. This copying approach generates diverse training data without requiring proportional increases in photographed image collection, significantly reducing time loss while maintaining inference accuracy.
Solution Approach 2:
The patent performs preliminary texture mapping and CG rendering to pre-generate diverse training images before the actual training process. By preparing comprehensive training data in advance through CG synthesis with mapped textures, the system reduces the need for extensive real-world data collection and processing during the training phase.
Data Source
AI summary
A training data generation device includes: a 3D model acquiring unit to acquire a 3D model of an object; a partial image acquiring unit to acquire a partial image that is an image area in which the object appears in a photographed image; a texture coordinate acquiring unit to acquire two-dimensional texture coordinates for texture-mapping the partial image on the 3D model on the basis of the partial image and the 3D model; a rendering condition acquiring unit to acquire a rendering condition that is a condition for rendering a 3D model with texture obtained by texture-mapping the partial image on the 3D model on the basis of the two-dimensional texture coordinates; a two-dimensional image acquiring unit to acquire a two-dimensional image by rendering the 3D model with texture on the basis of the rendering condition; and a training data output unit to output the two-dimensional image.