@SvenSommer I was talking to one of my co-workers about the facial recognition module, and he believes he can make it much faster. The issue, as he sees it, is not the actual recognition of whose face is in the image, but in all of the up-front work involved in preparing the comparison. First the image needs to be captured from the camera, then it has to be analyzed for the presence of one or more faces. Those parts of the image are cropped, resized and converted to grayscale, all on the CPU which does the operations one pixel at a time. The actual recognition of the face in the image is, according to him, the easy part and not the performance bottleneck.
He’s looking into whether any or all of this can be offloaded to the GPU and executed in ways that are far more efficient. Who knows… we might have good facial recognition working soon!
That said, I’ve found less and less practical use for separate display profiles, now having had my mirror on the wall for a few weeks. It turns out that most of the information on the mirror is useful to everyone in my household, so different profiles really only differ by one or two modules, and I’m not having problems fitting everything on to the screen at once. I would like the mirror to be able to greet me, however, when I stand in front of it, as it adds that cool factor. But I doubt I’m going to use it to actually change the display of the mirror.