Need an ADVICE on Xcode CoreML .mlmodel accuracy

I'm trying to edit Apple Flower recogniser example for recognising an artwork/paintings. It's working actually very well. Every time I point the camera to the artwork I trained before, it's always accurate. Problem is, it's showing > 98% confidence when I'm pointing to my table for example... Or almost anywhere where I point (not on artwork) it shows to match. You don't need to do whole project, just looking for advice. Apple example: [login to view URL] Text me, for more info. Pavol