Synthetic Dataset Generation

This document explains how WrenchML generates synthetic datasets for training computer vision models, the advantages of this approach, and best practices for optimizing your datasets.

What is Synthetic Data?

Synthetic data refers to artificially created data that mimics real-world data. In the context of WrenchML, synthetic data consists of computer-generated images of parts with simulated defects that closely resemble real manufacturing issues.

How WrenchML Generates Synthetic Data

The Rendering Pipeline

1. 3D Model Preparation

Your CAD model (converted to GLB format) serves as the foundation
Material properties are applied based on your configuration
The model is placed in a virtual environment

2. Defect Application

Selected defect types are applied to the 3D model
Geometric modifications create realistic surface imperfections
Position, size, and severity are controlled by your parameters

3. Scene Configuration

Virtual cameras are positioned to capture multiple views
Lighting is configured to match real-world inspection conditions
Background environments simulate factory floors or inspection areas