I am not sure if this is a Model Target issue, or a Magic Leap issue, but I wanted to post to see if anyone knows what the potential issue would be.
Using the Model Target Generator, I was able to create a model target, add a 360-view (with proper scaling and viewing bounds), and create the .unitypackage to import into my project. Using this untrained model target, the virtual model was properly overlaid on top of the physical object (while running on the Magic Leap).
Next, I created a new training set database, basing it off of the working untrained model target described above. It is processed, exported, and imported into Unity. However, after I switched the model target source to this trained dataset, the application can no longer recognize the physical object.
It's odd to me that a trained model wouldn't be able to recognize the object at least at the same basic level as the untrained model. Any thoughts on where to go from here?