Software detects food on photos and compiles individualized diet diaries

According to yesterday’s print edition of the Nikkei, Japan’s largest business newspaper, researchers from Tokyo University developed proprietary software that can detect and quantify different kinds of food captured on photographs.

The software analyzes colors and shapes to identify which type of food a person plans to eat. It’s also able to determine the size of bowls and plates to detect how much is going to be eaten. The software then classifies the food into different categories such as staple food, fruits, dairy products etc. For future reference, all of the data collected can be stored in the form of a daily diary .

The Nikkei says programming is near completion, with the researchers involved in the development stating the software’s error rate currently ranges between 10 and 20%.

The researchers aim at setting up a dedicated web site for their product by the end of this year. The service will most likely be offered for free.