-
-
Home Page
-
Scan Page
-
Account Prefences Page
-
Scan History Page
-
Inspiration
Food waste is a major global issue, with about one third of all food produced being discarded annually. In the United States particularly, about 30 to 40% of the food supply is wasted, totaling to over 100 billion pounds yearly. This food waste from both consumers and retailers contribute to climate change through methane emission from landfills. We saw food waste everyday in our lives as college students as we pass by the dining hall dish returns filled with plates that have plenty of food still on them. Also, many "expiration" dates are misleading, with food still being good for many days past the date.
What it does
Our website informs our users of different ways to assist them in reducing food waste through utilizing data from pictures of their fridges and pantries. It scans the pictures, identifying each food that appears, and logs it in a database. The user then gets informed of when their food expires and what recipes they could potentially make with the leftover food in their fridge and pantry, taking into account dietary restrictions.
How we built it
The frontend and backend were both coded utilizing Django. For the login system, we used the Auth0 API in order to implement sign-up and Google log-in. Our database was stored using SQLite. Our computer vision uses Gemini's object detection functionality to scan the images, and we also utilize Gemini API's access to Google Search Engine in order to search up expiration dates. Recipes are generated using Gemini API.
Challenges we ran into
One challenge that we ran into was generating the recipes to match with the data we got from the computer vision model. Our model returned specific brands of the food while the APIs we were trying to utilize did not use any brand names, and we also had to submit too many API requests, which made it take a while to load our website. This made it difficult to utilize these APIs. In the end, we had to streamline the process by prompting Gemini API to generate recipes using the database data.
Accomplishments that we're proud of
Our group is extremely proud that we were able to utilize computer vision when we have never used it before. We are also proud that our website is aesthetically pleasing and it conveys all the information that we want to after all the time we spent on UI/UX.
What we learned
We better learned how to evenly distribute work and plan towards a deadline. We improved our knowledge on Django and utilizing databases with SQLite. We also learned how to better prompt AI APIs to get better results. Finally, we better learned more uses for AI API, such as image recognition.
What's next for YumPie
As we continue to develop YumPie, there are several more features we would love to flesh out and make flourish. For one, we would add a function where it would tell you the best way to store each item, and if there is any way that you can improve based on what it sees in your fridge. We also wanted to add reminders either through push notifications or email notification to remind you a couple days before your foods expire. Similarly, we would keep track of how long each item has been in the fridge, and the recipe generator would prioritize ingredients that are closer to expiring.
Log in or sign up for Devpost to join the conversation.