Project: Guten Appetit
This is a project of the CIIS course (WS 2020/21). The content below was created by the project team: Kumar Abhishek, Vamsy Malladi, Maximo Schmidt.
Motivation & Idea:
Have you ever gone to the kitchen and wondered what could be cooked from the raw food items available? Or have you ever wanted to know which recipe could be the quickest? Or, ever wished to know what some strange looking fruit or vegetable at a supermarket is?
We have created a solution that mainly addresses these 3 questions. In addition to these, “Guten Appetit” also allows the user to have a digital copy of all the ingredients/food items they own, which can be used later while scaling-up the app to include features like, creating a shopping cart with items user does not have but needs to make his/her favourite recipes etc.
App and its usage:
1) User enters their favourite cuisines, food allergies if they have any and the kind of diet they follow.
2) User can populate the ingredients catalogue by either clicking a picture of the ingredient and letting our model predict the label, or by typing in the ingredient name.
3) Click the “Find recipes” button to get a list of recipes that were suggested by taking the user-preferences into consideration.
4) Users can save their favourite recipes and change the serving amounts if they choose to.
The image recognition component of our app is achieved through the fusion of a “Neural Network” and “LogMeal API”. We used “Transfer Learning” by retraining “MobilNetv2” model with “Fruits360”. The API was used to improve the predictions.
The recommendation process is done by integrating “Spoonacular API”. During this process, the ingredients selected from the “Digital Catalogue” and user preferences are considered. The recipes that the user sees here are from foodista.com
On top of these, the app also has its own custom camera and local database where user’s favourite recipes and their catalogue details are stored.
(The source of the recipes you see in the video and the pictures is foodista.com and licensed under a Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0/))