In the Binder Visualization tab you can import saved structures and computed volume fields which illustrate the binder identification process and its results.

Load Structure with Binder

Click Load *.gdt to load the structure of the identified binder labeled with a different Material ID.

In the 3D visualization, the distribution of binder between the grains is visible for the whole domain. In the 2D view you can evaluate the result of the binder identification in more detail, as seen in the example below (top: original structure, bottom: with labeled binder).

Load Confidence Field

During the identification the solid voxels are separated into two classes, binder and grain based on the selected Threshold. The confidence field contains the unsegmented result of the binder identification, which means a probability for each voxel to belong to one of the classes. Select Load *.npz to view this volume file in GeoDict.

Two fields are available, where the first field contains the probability for a solid voxel to belong to the first class (which is grain) and the second field shows the probability of a voxel to be binder.

For the visualization below, the second field channel_1:Numpy Field was loaded and you can observe easily where the neural network is confident of the choice and where it is uncertain: The dark blue areas are clearly marked as grain, the red areas are clearly marked as binder, and the values in between (bright blue to orange) are uncertain.