Training Neural Networks for Binder Identification in NMC Cathodes
PDF tutorial
The GeoDict AI module offers powerful AI-based functions for segmenting and analyzing 3D image data—such as from CT or FIB/SEM images—of complex, multiphase materials such as nonwovens, electrodes, or grain-binder structures.
In many cases, material phases cannot be separated conventionally due to identical gray value distributions. GeoDict-AI addresses this issue and enables shape-based separation using neural networks – for example, to detect binders, distinguish between different fiber types, or identify individual fibers.
In this PDF tutorial you will learn step-by-step:
- Inspect 3D structures and determine statistical parameters (e.g., diameter, orientation, volume fractions)
- Generate training data with structure-variable parameters via script (design of experiments)
- Train neural network, test, and validate segmentation performance
- Apply trained network for automatic material recognition in segmented scans
This tutorial is the ideal introduction to AI-supported image analysis with GeoDict and shows how segmentation can be automated, reproducible, and scalable.
Not registered yet? Register for free and enjoy all the benefits.
