Fridge Vision

Inspiration

Growing up, I would always be curious about what we were having for dinner. My parents often had to plan meals a week in advance, but sometimes they accidentally bought ingredients we already had in the fridge. Something as simple as deciding on dinner would lead to unnecessary stress and wasted food. This inspired me to create a solution.

What it does

Fridge Vision lets users snap a photo of their fridge, identify the visible ingredients, and suggest meal ideas based on what's available. Users can then select which ingredients they'd like to use, and the system, powered by Gemini, generates personalized meal ideas. This makes it easier to reduce food waste, save time, and plan meals more efficiently.

How I built it

I used Google’s Gemini AI to generate meal ideas based on selected ingredients. I also set up a basic web server to handle image uploads and process them using Gemini to return. After recognizing ingredients from the image, the server passes the ingredient list to Gemini to get recipe suggestions.

Challenges I ran into

One major challenge was figuring out how to process image files. At first, I tried converting the image into binary and passing it to Gemini directly, but that didn’t work—Gemini couldn’t interpret raw image data. I realized I needed to use an image recognition library (like OpenCV or Google Cloud Vision) to analyze the image and extract ingredient names first. Another challenge was setting up a secure, publicly accessible server. I had to learn how to properly configure endpoints and handle image data while keeping the server secure and reliable.

Accomplishments that I am proud of

• I successfully got Gemini to generate meaningful meal ideas based on real ingredients detected from fridge images.
• I learned how to process image data and integrate it with an AI API.
• I built a working prototype that bridges computer vision with generative AI in a real-world use case.

What I learned

This project taught me how to:

• Analyze and process image data to extract relevant objects (ingredients).
• Integrate a generative AI model (Gemini) with a web backend.
• Set up and secure a simple server for user interaction.
• Design a solution that addresses real-world problems like food waste and meal planning stress.

What's next for Fridge Vision

I’m excited to turn Fridge Vision into a mobile app so users can access it more easily. I also plan to add features like:

• Vegetarian and dietary filters (e.g., gluten-free, vegan).
• Skill level selection to tailor meal ideas to beginner or advanced cooks.
• Weekly meal prep suggestions to help users plan ahead based on what they have. Eventually, I’d like to integrate barcode scanning or voice input to make the app even more interactive and useful.

Built With

• css
• flask
• geminiapi
• html
• javascript
• python