The Modern Day Cookbook: Take A Photo and Grab the Recipe

Khryss | Published 2017-07-23 19:17

Gone are the days when you just stare at a food in the restaurant or Instagram and just silently wish you have the recipe.

When you look at a dish for the first time, can you easily recognize its ingredients? Well, if not, researchers at the Massachusetts Institute of Technology might've created some help for you. Just snap a photo and let their invention do the work!

They have made a machine learning algorithm that scans photos and predicts its ingredients.

Aptly named “Pic2Recipe”, Nick Hynes and his colleagues from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) trained this algorithm to recognize about one million recipes, each with a photo of the end result. When tested using a fresh photo, it was able to successfully guess 65 percent of the time!

"It would be really amazing to someday be able to take a photo of a dish you see in a restaurant and be able to figure out exactly how you can recreate it at home," Hynes told Motherboard. "I also imagine that people could use a tool like this to analyze their meals and determine its nutritional value. This would be particularly useful in restaurants and cafes when you don't really know what you're eating."

However, the AI still can't tell how the dish was prepared. It isn’t also very good in recognizing the subtleties of dishes. For example, when shown a picture of a butternut squash lasagna, it is more likely to show the recipe of just the generic lasagna. Identifying ingredients from foods like sushi and smoothies would also be a struggle as they’re made of blended or fine ingredients.

Nonetheless, the research team hopes to refine the AI so that it can analyse complicated meals and provide ways of preparing those meals. Though Hynes said that improvements might take a while. "Understanding recipes and their images is a tough problem that even Google research teams have difficulty with," he said. "I certainly think that [95-100 percent accuracy] is achievable, but it's going to require a much more in-depth understanding of the input."

Until then, the best way to know your food is to eat it. So eat up!

https://www.newscientist.com/article/2141363-ai-suggests-recipe-for-a-dish-just-by-studying-a-photo-of-it/

https://motherboard.vice.com/en_us/article/j5qnn7/mit-created-an-ai-that-knows-the-ingredients-in-your-food

Hey! Where are you going?? Subscribe!

Get weekly science updates in your inbox!