Unleash the Power of AI in GeoDict: Advancing Digital Material Development

In this presentation, we present an overview of the AI features capabilities included in the GeoDict 2024 release and an outlook into future developments. The main agenda topics include:

Overview & General Improvements of all AI usage scenarios currently implemented into GeoDict, for example in image segmentation in ImportGeo-Vol, in fiber and grain identification in FiberFind-AI and GrainFind-AI, and the possibilities to train your own neural networks using GeoDict AI. Multi-GPU support makes calculations faster and more efficient and FiberFind-AI is vastly improved.

Image Quality Enhancement addresses the latest techniques and methodologies in GeoDict to enhance visual data.

Multi-Channel Segmentation: Besides gray values, imaging devices may provide additional data (channels), which contribute to higher accuracy and better usability in the segmentation process.

Enhanced Fiber Identification Model: new pre-trained neuronal networks to largely improve fiber identification and fiber length distribution. The provided models are adapted to a wide range of different structure sizes, fiber orientations, and fiber diameters.

Nonwoven Digital Twin GeoApp: this newly introduced GeoApp automates the generation of a statistical Digital Twin of a nonwoven from a FiberFind-AI analysis result.

Artificial CT GeoApp uses structures generated in GeoDict to recreate an artificial CT scan.

This presentation provides a comprehensive overview of how AI is reshaping image processing and Digital Twin technology in GeoDict. The showcased applications show what is possible with GeoDict 2024, and offer a glimpse into the exciting possibilities that lie ahead in our future developments.