Track: Theme Track — Rethinking Waste
The waste problem we target (and how we'd measure it)
Households throw away fresh produce mainly because they cannot tell how long it will last and forget what is already in the fridge until it spoils. FoodIQ targets this household food waste: it estimates each item's remaining shelf life from a photo, gives storage advice that extends it, and keeps a fridge/pantry list with expiry warnings so food is eaten before it goes bad. We would measure success as the kilograms (or share) of tracked produce that ends up marked "used" rather than "spoiled/discarded" in the pantry log, compared over the first month against a baseline week with no reminders.
What it does
- FreshGuard (fresh produce AI inspector): snap or upload a photo and a PyTorch EfficientNet-B0 multi-task model identifies the item (10 classes: apples, bananas, bell peppers, carrots, cucumbers, mangoes, oranges, potatoes, strawberries, tomatoes), rates freshness (fresh / normal / rotten) and estimates the days of shelf life left.
- Storage and usage guidance: temperature, ethylene and packaging tips to make produce last longer, plus recipe and usage suggestions matched to ripeness, so nearly-ripe food gets used instead of thrown out.
- Smart fridge & pantry: tracks both fresh produce and packaged groceries with automated expiry warnings and a full history of scanned items.
- IngreScan-AI (packaged food scanner): reads ingredient labels with a vision LLM and flags additives, allergens and health risks against WHO/FSSAI data, with safer alternatives.
How we built it
React 19, TypeScript, Tailwind CSS 4 and shadcn/ui on the front end; an Express 4 / Node.js gateway with tRPC 11; a Flask (Python) AI inference microservice running the PyTorch/Torchvision EfficientNet-B0 model; SQLite by default (or MySQL/MariaDB) for per-user pantry and scan history; an OpenAI-compatible vision LLM for label reading.
Honest status
- The waste-reduction metric above has not been measured yet; no user study has been run.
- The produce model covers 10 types of produce only.
- FoodIQ grew out of Divyani's earlier food-safety project and was extended by our team; libraries and datasets are credited in the repository.
Code: https://github.com/madhavzanwar/tech4food Team: Madhav Zanwar, Divyani Papalkar
Built With
- efficientnet
- express.js
- flask
- node.js
- python
- pytorch
- react
- shadcn/ui
- sqlite
- tailwind-css
- trpc
- typescript
