|DoingLAB Founder and CEO Jin Song-baik speaks during an interview with The Korea Times, July 11. / Korea Times photo by Suh Jae-hoon|
By Kwak Yeon-soo
Jin Song-baik quit Samsung Electronics, where he worked as a software engineer, in 2014 to found a startup called “DoingLAB” in October 2016. DoingLAB uses an artificial intelligence (AI) engine to help users manage their weight effectively and work on eating more healthily.
Before putting energy into his own business, Jin worked as a researcher at Samsung Medical Center to study the dietary habits of diabetes patients and as a software engineer to develop a personal training app.
Both experiences have taught him more about the importance of eating behavior and food intake.
“I had an early idea while I was working on my project at Samsung’s C-Lab, which is an internal program for employees. But I also thought I needed real-time experiences,” Jin said.
According to Jin, the most frequent feedback he got from patients and app users was demand for nutritional counseling.
“There have always been apps in which users had to take note of everything they ate. However, these apps couldn’t work successfully in the long-term because users often forgot to type in details and keep their records,” he said.
“So I came up with an idea of digitizing the nutritionist to make everything seem easier.”
The company’s Diet Camera app uses the so-called “food lens” to recognize food images and save data of what the user is eating.
“The food lens identifies the food you’re eating automatically, saving you from entering it yourself,” Jin explained.
|An image of DoingLAB’s mobile app / Courtesy of DoingLAB|
The Diet Camera offers a huge food database, listing over 5,000 different types of food.
An effective calorie counter, it can take less than five minutes a day to track one’s dietary habits. It also offers an easy-to-use food diary for users to track what they eat.
“There are also nutritionists who’re keen to give advice on how best to proceed with diet plans,” he said.
However, Jin stated that there have been and still are many hurdles to tackle.
“We initially used Google images for our food database, but we realized that most of these nice and Instagrammable photos were no good for image recognition,” he said.
“We figured out that compiling ugly photos is a better idea because most food photos are just plain and simple.”
According to Jin, there are currently about 10,000 users that update photos of their food, and thanks to their rich data, the AI-powered Diet Camera can recognize more and more dishes.
The AI camera has learned to recognize photos of foods from current users. For the update, the engineers select photos and cluster old and new data together to develop patterns for identifying different types of foods.
Another feature of the food lens is that it can guess the food based on what it is surrounded by.
For instance, if an unidentifiable red sauce is located next to French fries, the lens recognizes it as ketchup instead of hot sauce.
“Studying the relationship of the food and its surrounding nature is important because one can infer from the background information,” Jin said.
However, he admitted that the company’s weakness is that it is hard to recognize what meat it is if it is cut into exactly the same shape.
“If it’s a chicken leg or lamb skewers, which are quite easy to identify, the AI-powered camera can definitely tell what it is,” he said. “However, if the meat is sliced into the same shape, it’s difficult to identify.”
Jin expressed that the key to a healthy diet is to know which parts of your meal are beneficial to your health and which are not.
“DoingLAB allows you to track the nutritional elements of your meal,” he said. “You can also track your calories and plan for losing weight or building muscles.”
Regarding criticisms that the app is unable to measure exact calories, Jin said that the app is intended to help users to be more generally aware of their habits.
“What human eyes cannot detect, the camera cannot detect either,” he said. “Since AI doesn’t have access to how much sugar or salt the food contains, it’s impossible to learn the exact calorie count.”
However, he stressed that the Diet Camera boasts 91 percent accuracy in identifying the foods so far.
“We’re still working to develop better AI engines to provide users with more essential information,” he said.
When asked about the greatest achievement so far, the startup founder mentioned winning 500 million won ($424,147) in funding from Naver, Korea’s web portal giant, in August 2018.
“Since then, we’ve also signed with various companies on application program interface (API). We’re still in discussions with a few healthcare and insurance companies, as well as medical centers.”
Although Jin started off as an app developer, he expects many more adventures ahead of him.
He said the company is willing to develop a more sophisticated food lens in order to provide in-depth and personalized nutritional counseling.
Apart from that, he showed his desire to become a food processing company, which provides healthy recipes and a wide range of foods.
“Our ultimate goal is to promote the importance of a healthy diet,” he said. “Since it’s a universal topic, we’re looking to expand our business to other regions as well.”
Share this post if you enjoyed! 🙂