It's common knowledge by now that food waste is a large issue, but we were shocked to find out 53% of food waste in the European Union comes from private households. This number comes from a major optimization problem that was solved decades ago at a corporation-level, but has traditionally been too small at a household level to bother: how do you efficiently use your food to avoid waste? Basically, how can you make a decentralized optimization problem simple enough so everyone can solve it at their own homes, reduce their waste and help the environment?
While we were going through this process, we realized how much additional potential this idea has. What if we can help people reduce their carbon footprint, become vegetarian or vegan, or avoid foods with allergens?
What it does
We built an AI-driven personal assistant that helps you save money, avoid food waste and live more sustainably. You can:
- Be recommended recipes that maximize ingredients currently in your fridge.
- Add your personal dietary choices for personalized recipes (e.g., vegetarian, vegan, sustainable, gluten-free).
- Add recipes you'd like to eat in the future to a grocery list.
- Take pictures of your pantry or fridge and keep track of your groceries without friction.
How we built it
Our frontend is a cross-platform Dart application, optimized for iOS use cases during the hackathon, but easily expandable to Android and web applications.
Our backend is a Flask server running on Python that works with a Firestore database and integrates with Azure Cognitive Services to read grocery store receipts, display a current view of your fridge, and recommend recipes based on your personal dietary choices & restrictions (e.g., vegetarian, gluten-free) while nudging people to more sustainable meals by displaying a CO2 score for every meal.
Challenges we ran into
Our product has a wide technological scope. We are curating personalized recipes, tracking what food items you have at home, let you build a cooking list which is in accordance with your dietary preferences and sustainability goals. With it come a lot of challenges in terms of matching inputs to data formats and creating a production-level user experience that seamlessly integrates our Flutter backend with the Python backend.
We worked on technology that lets you scan a grocery store receipt. This required us to integrate OCR technology into our application, which was a lot of work in terms of learning to understand receipts.
We integrated technology that lets you detect groceries, either in your fridge or in your pantry, and automatically . This required a custom training a custom computer vision model on Azure, as well as building out object detection technology with OpenCV.
Accomplishments that we're proud of
We're proud of creating an MVP for a smart fridge camera that leverages machine learning & computer vision to detect new groceries you put in your fridge and shows the view from your fridge on our app, including boxes that highlight which groceries are in the fridge.
What we learned
We learned it's extremely difficult to work with machine learning and computer vision, but still worth it due to the massive value add to our users. We understand now the difficulties of integrating the challenges of building for the environment and adding value for our users.
What's next for Nevera
We want to work with smart fridge manufacturers to integrate our technology to make it easier for their clients to live sustainably, as well as work further on our machine learning models to recognize all types of food one could have in their kitchen or pantry. We will add functionality to track expiration dates. We will add internationalization support so Nevera can work with all cuisines. We will add support for more dietary choices & restrictions, as our goal is to be as inclusive as possible in the long term.
Log in or sign up for Devpost to join the conversation.