Generally I think modeling in the computer without proper detail in measurements and geometry is error prone. I will normally capture a 3D point cloud (can use a Kinect for this) and perform a geometry scaling, then use AI to map it into 3D surfaces compatible with CAD. This will take 1-2 days to model. Then I manually model all the elements in minute and exactly detail. Finally, after all the work is done, I assess the coloring of the cabinets to see how I feel about green vs. chartreuse paint, on a high-end monitor typically used by radiologists to assess human disease. Some people ask why not just hold a color ribbon up to it to get an idea, I say that's a waste of time, not enough.