AutoChef: Computer Vision for Automated Ingredient-to-Recipe Matching
Date:
[Paper] [Presentation] [GitHub]
Abstract
When many of us go grocery shopping to our favorite convenience store, we have an idea of what we want to cook for the week and shop for the required ingredients accordingly. However, after cooking our intended meals for the week, more often than not, we have to deal with leftover ingredients. Often, these leftover ingredients sit in the fridge until they are ultimately thrown out, leading to food waste. These leftovers are tedious to identify one-at-a-time in the fridge, and sometimes, are completely unrelated: this week, Sharan - one of our team members - was left with kale, garbanzo beans, kimchi, lamb chops, and tuna. Moreover, this act of food waste reaches further than simply inconveniencing individuals - it is also a nationwide, environmental issue: Feeding America estimates that nearly 40% of all food in the U.S. is thrown out.
To help users solve this bothersome, yet relevant problem in their cooking lives, we developed AutoChef: a web application that automatically identifies leftover ingredients and recommends recipes that maximize usage of leftover ingredients. Unlike pre-existing recipe APIs such as SuperCook, which require users to manually figure out and type in leftover ingredients, AutoChef quickly identifies multiple ingredients from a single photo, allows users to adjust the identified list of ingredients, and recommends recipes based on cuisine, dietary restrictions, type of dish, and intolerances. Furthermore, AutoChef acts as a one-stop-shop for all cooking-related needs by also allowing users to favorite recipes and providing detailed recipe instructions.