My problem is, taking pictures is more effort than picking an item off a list, as current caloric counters do.
Most people eat roughly the same things on a regular basis, so they'll end up ticking away a meal before/after you have it and be done with it, with this, you'd be taking pictures while having your meal.
another use case is planning a day ahead, again, pics don't work here, can't take pics ahead of time.
And of course, the result from the pics have to be corrected, so the app learns, it seems easier to just get the item from a list immediately, without having to take a pic first, auto completion on an input works wonders (though of course you'd pick from a longer list)
Maybe i'm just old or not enough of a techie, or photographer, but for me typing a short text, even on a phone, is actually faster than taking a pic.