High Quality Sythetic
Defect Data Generator for Industry
AIDG is an intelligent sample generation software powered by generative AI, addressing the challenge of insufficient defect data in inspection model training and validation. It helps users to efficiently obtain high-quality samples and high-performance AI models, while saving time, effort, and cost.
Product Highlights
High Fidelity
Built on StableDiffusion framework and Aqrose’s deep expertise in industrial defect inspection, AI algorithms in AIDG adapt to various workpiece structures and backgrounds, generating defects with authentic edges, textures, and color details.
Fast & Flexible
Supports multiple generation modes: fine annotation, parameter-based auto generation, batch generation, and mixed-defect generation. This enables rapid creation of high-quality, diverse, and balanced samples.
Defect Library
Allows users to accumulate and manage defect materials, making it possible to transfer and reuse defect information from one product to another, which is highly valuable in scenarios with production line changeovers.
How It Works
Import Real Defect Samples
Build A Defect Library with Annotations
Import OK (Non-Defect) Samples
Set Generation Requirements
Run AI Generation or Migration
Review and Export Results
Application Scenarios
High Yield Rate Production
When defects are rare, collecting enough samples takes too long. AIDG generates synthetic defects for training and validation, ensuring faster, higher-performing model development.
Low-Volume Production
In industries where daily output is limited, AIDG accelerates dataset creation, making AI training and testing feasible.
Emerging Defect Types
For emerging defects that lead to missed detection, AIDG can generate batches of synthetic images from just a few real samples, supporting rapid model upgrades.
Production Line Changeovers
During product variant changes, AIDG generates simulated defect data by transferring defects from existing products, accelerating AI model deployment for the updated production line.
