How much do you love ramen? Can you just look at a picture of a ramen bowl and tell which restaurant it came from? If your answer is yes, it’s likely we’ve got a bigger fan than you. A recent study shows that machine learning software can not only identify the exact shop of a ramen bowl, but also the franchise of the same shop.
This delicious research was done by Kenji Doi, who used Google’s AutoML Vision and machine learning model to classify every menu from Ramen Jiro, the most popular restaurant franchise in Japan. He fed a dataset of 48,000 photos taken from its 41 shops. It sounds quite crazy when all of its franchises serve the same basic menu, and they all look like the below photo.
However, for Google’s AutoML Vision, the job wasn’t that difficult. When Kenji uploaded those photos of ramen bowl to AutoML Vision, it took 24 hours to finish training the data, and the model could predict the exact shop the ramen came from with 95 percent accuracy.
Earlier this year, Google unveiled its Cloud AutoML software to take the pain out of AI coding from the developers. It allows the users to develop machine learning software through a simplified, drag-and-drop process. To identify a specific product, in this case, ramen, AutoML Vision develops customized ML models automatically. All you need to do is upload a set of well-labeled images and clock a button. Kenji believes his model is accurate enough not only to distinguish between the shape and cut of the bowls but also to identify the cuts of meat and placement of toppings of each shop. It looks like the days are not far when you could find the location of the restaurant from a tag-less ramen on your Instagram using this software.