FridgeWhiz

Inspiration

The project was inspired by the challenge of making healthy eating more accessible and reducing food waste. Many people struggle with meal planning, especially when they don’t know what they can make with the ingredients they already have. We wanted to create a tool that simplifies meal preparation and encourages healthy choices, all while reducing unnecessary grocery trips and food waste.

What it does

Our project allows users to take a picture of the inside of their refrigerator. Using image recognition, the app identifies the ingredients present and generates recipes based on what is available. It also recommends a shopping list of healthy food options that can complement the existing ingredients and suggests healthy recipes tailored to the user's preferences.

How we built it

Image Recognition: We used the YOLOv8 model for object detection to identify food items in the fridge. YOLOv8’s speed and accuracy make it ideal for identifying various ingredients from a single image. Machine Learning & ChatGPT: We integrated machine learning algorithms to help recommend personalized recipes based on the identified ingredients. We also used the ChatGPT API to provide dynamic and natural recipe explanations, answer user questions, and give food-related advice. Frontend: The website was built using HTML and CSS for a clean and responsive user interface. Backend: We used Flask and Python to handle the image processing, run the machine learning models, and manage data flow between the frontend and backend systems.

Challenges we ran into

Accuracy of ingredient detection: One of the biggest challenges was training the YOLOv8 model to accurately recognize food items in varying conditions, such as different lighting, packaging, and camera angles. Recipe generation: We had to ensure the recipe suggestions were realistic based on limited ingredients and user preferences, such as dietary restrictions. Integrating the ChatGPT API for seamless recipe explanation and suggestions also required optimization to maintain fast response times. Website design: Creating a user-friendly interface while maintaining smooth functionality across devices took several iterations of design and testing.

Accomplishments that we're proud of

Successfully implementing the YOLOv8 model to accurately identify a wide range of ingredients from refrigerator photos. Seamlessly integrating the ChatGPT API to generate personalized, healthy recipes based on available ingredients and to provide an interactive experience for users. Building a functional and clean website interface using HTML and CSS, ensuring ease of use for users across different devices.

What we learned

We learned a lot about computer vision, particularly through working with YOLOv8 for object detection and optimizing its performance for real-world scenarios. Using the ChatGPT API taught us about creating a more interactive and conversational user experience. We also improved our skills in web development by building a responsive interface using HTML and CSS. Integrating multiple technologies to create a cohesive product gave us experience in managing backend processes, frontend design, and API connections efficiently.

What's next for FridgeWhiz

Expand Ingredient Recognition: We plan to improve the YOLOv8 model to recognize a broader range of ingredients, especially those common in international cuisines. More Personalized Recipes: Adding more filters for dietary preferences (such as vegan, keto, or gluten-free) and allowing users to input their food preferences will make the recommendations even more tailored. Mobile App: We’re planning to develop a mobile app for easier fridge scanning and recipe viewing. Grocery Store Integration: We aim to integrate the shopping list feature with local grocery stores, allowing users to check availability and prices in real time.