GenAI – quality control with artificial image data

© Fraunhofer IPA

In industrial quality control, accurate and reliable image datasets are crucial when it comes to training AI systems. Capturing this data in the traditional way normally involves a great deal of time and money, particularly if defects, flaws or certain configurations are only a rare occurrence.
 

Value proposition

You can reduce the time and expenditure involved in data capture considerably by using synthetic image datasets. They allow you to simulate various component properties, such as geometry, material and surface finish, as well as realistic lighting conditions and optical effects for different scenarios. With an AI-based defect generator, we create fault images automatically with corresponding annotation as an optimal way to train the AI models.
 

Investment and scope of services

With a proof of concept, we can offer the following services, starting from 20,000 euros:

  • Definition of the necessary image data and its variants to train an AI model (e.g., defect detection on component surfaces)
  • Generation of synthetic data (e.g., with GANs or diffusion models)
  • Evaluation of models with synthetic training data to validate the application in practice
  • Implementation of an AI application as a proof of concept for quality control

Contact us to develop and improve your AI-based quality control by generating artificial image data in a low-cost process.