FridgeFix

• 

  Two ways to start

• 

  Ingredients and preferences

• 

  Personalized recipe matches

• 

  smart shopping list

• 

  On-device food recognition

• 

  Nutrition and healthier swaps

• 

  meal ratings and history

Inspiration

Students often want healthier meals but must work with limited ingredients, time, and money. FridgeFix turns those constraints into practical recommendations.

What it does

Users can enter available ingredients and preferences to receive ranked recipes, missing-item shopping lists, and cooking instructions. They can also upload a meal photo, confirm the detected foods and portions, then compare healthier swaps and nutrition estimates.

How we built it

We built FridgeFix with HTML, CSS, and JavaScript. A pretrained ONNX Food-101 model runs through Transformers.js in the browser, while ratings, leftovers, shopping items, and meal history use local storage.

Challenges we ran into

Food-101 recognizes complete dishes rather than every separate item, hidden ingredient, or exact portion. We therefore require users to confirm predictions and choose approximate serving sizes.

Accomplishments that we're proud of

We combined meal recommendations, allergy filtering, recipes, estimated grocery prices, nutrition comparisons, ratings, and history in one working website.

What we learned

AI predictions should support human decisions rather than replace them. Clear limitations and user confirmation make the results more trustworthy.

What's next for FridgeFix

We would expand the recipe and nutrition databases, improve multi-item food detection, and personalize rankings using previous ratings and meal history.

Built With

• css3
• github
• html
• hugging-face
• javascript
• machine-learning