Building a classifier and applying to live data
We can design our classifier in the familiar way. We define classes, paint labels and build a model. The plastic samples, included with each perClass Mira Stage, are marked with a letter, followed by a number. The letter describes the material, and the number its variant. In our example, we want to distinguish different plastic materials, irrespective of color. Therefore, we define background and material classes A - E
Then, we pain the labels and create a model by pressing Model search
We can see that all different variants of material A can be separated from different white plastics B-E. This demonstrates the unique value of spectral imaging where we may base our interpretation on material composition, not appearance.
In order to demonstrate how our solution works live, we assign the Cycle scanning area command to the button A in the Stage panel. By pressing A button of the stage, the software switches to the Camera mode and stage starts cycling.
In the toolbar, we may now see also the Decisions button enabled, in addition to Raw and Corrected buttons. We may choose how do we want to see the live data stream.
The Camera panel will then also show a classifier speed (the red line) .
This concludes our acquisition tutorial. We have learned how to acquire references, record scans, define models and run the full correction and modeling pipeline on a live data stream.