in #CASE you missed it… New System Can Translate Food Pics Into Recipes
MIT researchers have developed a system that will find recipes based on food pictures
Researchers at MIT have developed a neural network called Pic2Recipe that can take a picture of a dish and work backwards to develop a recipe for it
The researchers developed a database, “Recipe1M” from millions of recipes and pictures on sites
So far, the network works well with baked goods, but has problems with more ambiguous food, as well as foods that have multiple recipes
Have you ever eaten a dish and instantly wished you had the recipe? Well, a recently developed system may soon make that happen, simply by snapping a picture of your food.
Researchers at MIT have developed a neural network called Pic2Recipe that will be to food what Shazam is for music. You take a picture, and it work backwards to develop a recipe for it. The researchers developed a database, “Recipe1M” from millions of recipes and millions of pictures on sites, and were able to train the network to “find patterns and make connections between the food images and the corresponding ingredients and recipes."
As for how well it works, it seems there’s still some work that needs to be done. The MIT team calls it a work in progress because it does well with baked goods. However, it's not up to snuff when “identifying more ambiguous foods, like sushi rolls and smoothies.” There’s also a bit of confusion when there are multiple recipes for the one dish