Instagramming pictures of your food at dinner could finally have a purpose, thanks to the wonders of science.
Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory have developed a program that will analyze your food photos and scour it’s database of over one million recipes for a match.
To build the program, dubbed Pic2Recipe, the team first had to build a database of recipes. They combed through cooking websites and uploaded over one million recipes before training a neural network to connect the images of food to corresponding ingredients and recipes.
The database is now the largest, publicly available collection of recipe data.
Javier Marin, an MIT researcher and one of the co-authors of the project, said that the system was needed, “in order to better understand people's eating habits and dietary preferences.”
“When it comes to food, there was not any large-scale dataset available in the research community — until now,” Marin said.
Theoretically, you could use it to find out how to make your favorite dish, but the team behind the project had more practical ideas.
“This could potentially help people figure out what’s in their food when they don’t have explicit nutritional information,” project lead Nick Hynes said in a press release.
Marin said the team foresees many other potential applications for the project, from estimating nutritional information to food manipulation.
Food manipulation, Marin explains, would entail having the system come up with alternate ingredients to produce a healthier meal than the original recipe.
The team has set-up a demo website where foodies can put the system to the test with their own pictures, or try it out using random food images.
The system works particularly well with desserts, which make up 22 percent of the recipe database.
The team, composed of researchers from MIT, Qatar Computing Research Institute and Polytechnic University of Catalonia, will present the project at the Computer Vision and Pattern Recognition conference in Honolulu later this month.
Javier Marin, MIT researcher and co-author of the Pic2Recipe project