VISION - Versatile AI systems for intelligent and optimized NDT processes

The project VISION aims to fundamentally transform the use of Artificial Intelligence (AI) in the field of Non-Destructive Testing (NDT). At its core is the adaptation and advancement of Foundation Models—pre-trained AI models that are tailored to specific NDT applications through Self-Supervised Learning and targeted fine tuning. This innovative approach significantly reduces the need for large amounts of labeled data while still enabling the development of powerful and cost-effective AI models. The project activities involve the comprehensive curation and preparation of large-scale NDT data, followed by the self-supervised training of Foundation Models. These models are then tested for their suitability and performance through various downstream tasks and feature-based evaluation methods. The primary objective is to optimize the quality and efficiency of these models, making them suitable for demanding industrial applications such as the inspection of high-tech components or the enhancement of production processes. A key aspect of the project is ensuring the quality and trustworthiness of the developed models, known as "Trusted AI." This not only aims to increase the acceptance of AI solutions but also to strengthen the competitiveness of Austrian and European companies. VISION thus contributes significantly to Europe