Unsure about your vegetables or fruits freshness? This app allows you to get a second opinion on the state of your veggies! Built for the 2026 Hack For Humanity Competition. This project uses MobileNetV3 as a base model, then further trained to recognize the mold on vegetables and fruits. Using ONNX to export the model into a C# backend for quicker loading and access times.
Food waste is a major global issue as a whole humanity throws out 1+ Billion tons of edible food each year. As food is thrown out just because it looks questionable (e.g bruising or looks cheap) this leads to:
- Unnecessary food waste
- Increased environment impact
- Higher grocery costs
VeggieWaste uses machine learning to analyze images of vegetables to determine whether they are fresh or rotten, helping users make educated decisions.
- Image upload via browser
- Freshness classification via AI
- Confidence scoring for transparency
- ONNX model inference in the backend
- React + Typescript
- Vite
- TailwindCSS
- Axios
- ASP.NET Core Web API
- ONNX runtime
- Imagesharp
- MobileNetV3 Convolutional Neural Network
- Light weight and high accuracy-to-performance ratio
- Exported to ONNX for cross-platform inference
- Real-time image classification
- The user sends an image through the web interface, where it is then sent to the backend.
- The backend receives the image then:
- Preprocesses the image
- Runs inference using MobileNetV3-based ONNX model that was trained using 10,000+ images.
- The API returns the classification and confidence score.
- The frontend displays the result instantly.
- Node.js
- .NET 10+
- Git
dotnet runRuns on: A local host of your choice.
npm install
npm run devRuns on: A local host of your choice.
The dataset used in this project is from Kaggle made by Swoyam Nayak.
- Insomnia for API Testing
- Ngrok for LAN testing across devices
- Hack For Humanity for hosting this hackathon