Alper Yilmaz
Professor at Ohio State University, where my research is in computer vision. Founder of PIQ Labs, which makes Nutrician.
My academic work is on how machines interpret images — how a system decides what it is looking at, how confident it is entitled to be, and what happens at the edges of that. That is the same question a food-photo app answers every time somebody photographs a plate, and the answer is less certain than the interface usually admits.
Nutrician exists because of the gap between those two things. Recognising that a photo contains chicken and rice is a problem largely solved. Deciding how many grams are on the plate is not, and no amount of model size fixes it, because the information is not in the image. So the app is built to say which of the two it is doing: every figure carries its origin, and an estimate is shown as a range rather than a number that looks like a measurement.
Disclosure
I have an obvious commercial interest in this subject. PIQ Labs LLC sells Nutrician, a paid subscription product, and it competes with the apps I write about. Where I have cited published research, I have tried to cite what it actually found rather than the part that suits me — including where it is inconvenient, and including the limits of our own internal testing.
Elsewhere
Writing
Guessing looks exactly like knowing — food-photo apps identify meals well; the numbers they attach are a different problem, and the interface does not tell you which is which.