USPTO Patent Grant: Machine Learning Method on Vectorized 3D Models

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted a patent (US12586309B2) for a machine learning method on vectorized 3D models and a learning system. The patent, assigned to Vizuro TAIWAN COMPANY LTD., details a process for training AI models using synthetic images rendered from vectorized 3D models.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has issued patent US12586309B2, titled "Machine-learning method on vectorized three-dimensional model and learning system thereof." The patent, granted to Vizuro TAIWAN COMPANY LTD., describes a method for training artificial intelligence models using synthetic images rendered from vectorized three-dimensional models. The system acquires engineering design drawings or stereoscopic scan files, converts them into a specific file format, and uses a rendering program to create a dataset of diversified synthetic images for training and evaluating machine learning algorithms.

This patent grant is primarily an intellectual property matter and does not impose new regulatory obligations on businesses. However, companies involved in AI development, 3D modeling, computer graphics, or manufacturing may wish to review the patent's claims to understand the scope of the protected technology. The filing date was January 26, 2024, and the grant date is March 24, 2026. No compliance actions are required for regulated entities based on this patent grant.

Source document (simplified)

← USPTO Patent Grants

Machine-learning method on vectorized three-dimensional model and learning system thereof

Grant US12586309B2 Kind: B2 Mar 24, 2026

Assignee

Vizuro TAIWAN COMPANY LTD.

Inventors

Ting-Yuan Wang, Wei-Chin Liu

Abstract

A machine learning method on vectorized three-dimensional models and a learning system thereof are provided. The learning system uses a computer system to perform the machine learning method using synthetic images rendered from vectorized three-dimensional models. In the method, the system acquires a computerized engineering design drawing or a stereoscopic scan file of a target object. The engineering design drawing or a stereoscopic scan file is then converted into a specific file that can be processed to render synthetic images by a rendering program. A great amount of diversified synthetic images forms a training and testing dataset for an ensemble of machine learning algorithms to train an artificial intelligence model and evaluate its accuracy.