Josephus67/Ghana-Hack-AI

Crop Guard . # Live Demo. youtu.be ## Architecture Diagram Below is a high‑level flow of how Crop Guard works, from user action in the app to model inference and back: ## Screenshots Screen Preview Home Image-Picker Camera Preview History Analysis ## Table of Contents 1. Project Overview 2. Table of Contents 3. Deployment & Next Steps 4. Performance Metrics 5. Future Work 6. Acknowledgments ## Crop Guard **AI‑Powered Crop Disease Detection** Crop Guard is a React Native mobile app that helps Ghanaian farmers—especially those just starting out—identify diseases in key crops (maize, cassava, cashew, tomato) using their smartphone camera. By streamlining disease diagnosis, it promotes early intervention and supports Ghana’s agricultural development. **Challenge Tackled:** Smart Agriculture – Crop Disease Detection **Tech Stack:** - **Frontend:** Expo / React Native (see `package.json`) - **Backend:** Python 3 + FastAPI (see `requirements.txt`) - **ML Framework:** TensorFlow 2 (Xception fine‑tuned models) - **Other:** ngrok for local tunneling, CORS enabled **Folder Structure** ``` Ghana-Hack-AI/ ├─ crop-guard-frontend/ # Expo app │ └─ … # React Native code, assets, package.json ├─ crop-guard-backend/ # FastAPI server │ └─ … # API routes, model loader, requirements.txt ├─ crop-disease-models/ # Trained Xception models & notebooks │ └─ … # .keras files, Jupyter notebooks └─ README.md ``` --- ## Deployment & Next Steps ### Local Testing with ngrok We’re using **ngrok** for local testing due to model size constraints: 1. **Start FastAPI server** ```bash cd crop-guard-backend/ uvicorn app.main:app --reload --host 0.0.0.0 --port 8000 ``` 2. **Launch ngrok** ```bash ngrok http 8000 ``` ### Considered Render Deployment - We included a `render.yaml` in the repo, but models currently exceed free-tier sto …