Method and Systems for Providing Synthetic Labelled Training Data Sets and Use of Same

Simple SummaryContent extracted from patent full text and abstract with AI.

This invention describes a computer-implemented method and system for generating synthetic, labeled training datasets by utilizing CAD models of objects. The process involves automatically selecting sub-objects within a CAD model, generating multiple rendered images containing these sub-objects, and labeling the images based on the CAD information to build comprehensive training datasets. These datasets can then be used to train machine learning models for tasks such as classification and object detection in images, potentially reducing the need for manual labeling.

Use CasesContent extracted from patent full text and abstract with AI.

• Training machine learning models for automatic visual inspection in manufacturing (e.g., printed circuit board assembly, automotive production).
• Developing and validating image recognition systems for quality control and defect detection in industrial processes.
• Accelerating AI development in robotics for object recognition and manipulation tasks.
• Training autonomous vehicle or train perception systems using CAD models of environments/scenes.
• Supporting inventory tracking or picking systems in warehouses where CAD models of items are available.
• Medical imaging analysis and training where CAD models of anatomy or medical devices are used.
• Any other domain where synthetic labeled images derived from CAD models can replace or supplement real-world data collection for AI training.

BenefitsContent extracted from patent full text and abstract with AI.

• Drastically reduces time and cost required for generating large labeled training datasets, especially compared to manual labeling.
• Minimizes labeling errors and increases dataset accuracy through automated, CAD-driven labeling.
• Enables the generation of diverse image conditions (different angles, lighting, backgrounds, occlusion, etc.), improving the robustness of trained AI models.
• Supports training for rare or hard-to-capture scenarios that may be underrepresented in real-world datasets.
• Allows efficient adaptation to new products or parts by simply updating CAD models rather than capturing new photos and labels.
• Scalable for industrial and commercial applications where rapid AI deployment and continual adaptation are needed.
• Facilitates advanced quality control, predictive maintenance, and process automation initiatives across multiple industries.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

The invention relates to a computer-implemented method for providing a labelled training dataset, wherein - at least one sub-object (21, 22, 23, 24, 25, 26) is selected in a CAD model (1) of an object (2) comprising a plurality of sub-objects, - a plurality of different render images (45, 46, 47, 48) is generated, wherein the different render images (45, 46, 47, 48) contain the at least one selected sub-object (21, 22, 23, 24, 25, 26), - the different render images are labelled on the basis of the CAD model to provide a training dataset based on the labelled render images (49).