Fridg-e

Have you ever found something in the fridge that you completely forgot about? Well, we have!

Sometimes, while digging through our fridges, we'd find really, really expired food. Cream cheese, 4 years past expiration. Condensed milk, hidden all the way back, somehow crystallized. Food would constantly expire. Mold would appear randomly. It was a terrible habit, and we wasted. so. much. food.

So we decided to tackle this common household problem. Food waste is no more.

Our solution is Fridg-e. He'll help you track the items in your fridge. He'll make sure you don't waste any more food.

But what does Fridg-e do?

Displays what is in your fridge right now
Automate tracking expiration dates, calories, carbon footprint, and more!
Computer vision to automate tracking items in your fridge
Comes up with recipes for you to use (prioritizing things that might expire soon)
Sorts your fridge items in various categories!

How we built it

We placed a camera inside the fridge to continuously stream the video to our server (in this case, our laptop).

Now, to recognize food, we used YOLOv11. By placing it on the door, it sees only what is put in and taken out. It detects general items that people commonly put into fridges such as apples, bananas, etc. Though, we also implemented brand specific recognition for things such as drinks (as a cool proof of concept that we could scale upward)

Something that took quite a bit of effort was taking into account when the fridge opens or closes. Imagine you're the camera. And the fridge door closes. You're gonna see everything in the fridge, and upload everything onto the database.

We don't want that. To see when the fridge is closed, we measured the brightness of the image. If it's not bright enough, then the fridge is closed.

But that still doesn't solve the issue of seeing everything in the fridge. In the motion of closing the door, there's still light, so we implemented a buffer. Now if the door is closed, the last second of data is deleted before it's uploaded.

For our website, we used Firebase for both authentication and our database. On the dashboard, you can see the food, calories, carbon footprint, and predicted expiration date. Finally, Gemini generates a recipe using the foods closest to expiration.

We also rendered Fridg-e! He's a 3D model (rendered with threejs) that shows what you currently have in the fridge, and whether it's open or not.

We made an actual thing that works, and something we can use personally! We were able to tie the computer vision and frontend together, while still having a quick response.