Sound impossible? We thought so, too. But on a test run, the app [iTunes link] accurately identified everything from a bag of chips to barbecue ribs. For each photo, it returns a ballpark calorie range for one serving of the item pictured.
Hardcore counters, beware: The app’s calorie counts don’t account for serving size and seem to be averages for the food category rather than a calculation of the specific food item (i.e. “turkey sandwich” rather than “turkey sandwich with one slice of cheese and mayonnaise”). They are, however, generally similar to the range that might be included in a calorie counting reference book.
DailyBurn CEO Andy Smith is mu m on how the app actually works, referring to it the same way that the Meal Snap website does: “Magic.” He does say that there are both humans and algorithms in the secret sauce and that DailyBurn’s experience building more old-fashioned calorie counting apps helped provide information that makes the photo app possible. Somehow the company seems to have paired an optical recognition system with its existing food database.
The app doesn’t always identify food correctly, and sometimes calorie counts are too broad to be useful. But users can rate accuracy of each identification and calorie count, ostensibly helping to improve the optical recognition component. Manually labeling food photos also makes the results more accurate.
But accuracy isn’t the point, Smith says. Intentionally keeping a record of the foods that one eats is the biggest advantage of calorie counting. With the Meal Snap app, users automatically collect a visual food d iary as they count...
