A team of researchers at the Massachusetts Institute of Technology tried to create a deep learning algorithm that could predict a recipe just based on a photo. It is called Pic2Recipe and was the result of a research published on the 20th of July.
Just upload a photo of a dish and receive, within seconds, the complete recipe. A very interesting app that could pave the way for new uses even if it is just the start.
How does it work?
After inserting the picture of a dish to the app, Pic2Recipe provides a list of up to five recipes sorted according to the confidence the system places in the various alternatives. At this time, the database is based on American recipes.
Using those recipes and the associated images, the team was able to train the software to use object recognition to pick up on what each dish’s ingredients might be. With a list of ingredients, the system then selected the recipe that best matched the list. Pic2Recipe was able to recognize ingredients like flour, eggs, and butter.
The system has been developed using artificial intelligence and a database of over one million recipes. Although it's still a prototype, this algorithm has a 65% success rate. However, the program still has a few gaps. The program has trouble with dishes that are a bit more ambiguous, such as smoothies. Similar recipes with a number of different variations, like lasagna for example, also tended to confuse the program.
Watch the video below to see how the future might be: