"Im2Calories" which can calculate calories from photographs of meals and dishes with Google's artificial intelligence

BySea turtle

Google is focused on developing artificial intelligence (AI)Unattended driving carYaDroneWe are promoting various projects using artificial intelligence such as AI which automatically calculates calories by analyzing photos of food "Im2Calories"Was newly announced.

Google's A.I. Is Training Itself to Count Calories In Food Photos | Popular Science

Google announces a new AI project "Im2Calories" - Calories Calculated from Food Photo - CNET Japan

It took place in Boston at the end of May 2015Rework Deep Learning SummitAt Google's researcherKevin MurphyA new project called "Calculate calories by analyzing food reflected in photographs using elaborate deep learning algorithms"Im2Calories"Was announced. We acquired this project for 50 billion yen in 2014A mysterious artificial intelligence development startup company "DeepMind"It is said that it utilizes the technology of.

Google acquired 50 billion yen of a mysterious artificial intelligence development startup company "DeepMind" - GIGAZINE

Im2Calories scans photos of dishes on a dish, for example, analyzes the type, number and size of the food reflected in the photograph, such as "2 eggs, 3 pancakes, 2 bacon" A system that can be done. It seems that we can also distinguish the seasoning used in cooking. Also, Im2Calories can be scanned even if it is not a high-resolution photograph, and it is enough if enough image quality is uploaded to Instagram.

ByJessica Spengler

According to Mr. Murphy, "Using Im2Calories makes it easier to attach everyday meal records, write down foods one by one, and eliminate the tedious task of estimating calories according to the size of the dish." For example, when Im2Calories erroneously recognizes the photo of egg-fried eggs as poached eggs, we aim to construct a "semi-automated system" that allows you to change the name of the dish manually.

ByU. S. Department of Agriculture

Im2Calories seems to measure the depth of each pixel of the photograph, perform a visual analysis, combine it with pattern recognition, and associate the analysis result with huge calorie information to calculate calorie content. "There is an error of about 20% in the calorie calculation result at the present time, but from now on, we plan to experiment with the system by having a large number of subjects cooperate, and the goal is to make the error almost zero after one year "Mr. Murphy talks about future prospects.

The details of the service and the date of launch are not yet clear,Obesity super powers AmericaSo it seems to be a service that seems particularly useful.

in Web Service,   Junk Food, Posted by darkhorse_log