Remember when “Julie and Julia” was the craziest thing to ever happen to the food world? We’ve come a long way, baby.
Now, there’s an app that claims to actually look at photos of food and know what the dish is and even how to prepare it.
Pic2Recipe is the artificial intelligence system developed by researchers at MITs Computer Science and Artificial Intelligence Laboratory that analyzes photos of food and actually tells you what the food is and how to cook it.
The researchers used websites like All Recipes to create the Recipe 1M database with you guessed it over one million annotated recipes, complete information about various ingredients. The researchers then “used that data to train a neural network to find patterns and make connections between the food images and the corresponding ingredients and recipes.”
Apparently the system is already well versed with cookies and baked goods that make up a large part of the database, but less so with items like sushi and smoothies. The system can also be stumped by “similar recipes for the same dishes.” When we put it to the test, using personal photos from meals at different restaurants, here were our results:
The program successfully identified this dish as “Eggplant Pasta,” though the accompanying recipe skipped any cheese and the first suggestion was pasta with sausage.
This photo of eel sushi was less successful and had zero hits. Clam chowder, tacos and Caesar salad suffered the same sad fate.
The program bounced back with several chicken finger recipes for this photo.
This carrot cake photo was a mixed bag, as it brought up recipes for vanilla frosting but also ice cream cake.
The program has some kinks to work out since there were several foods it could not recognize, but it has potential. The researchers see this as a potential tool to help people figure out the nutritional content of their meals, recreate restaurant experiences at home and – of course – up their social media + food obsession.