salaudeenabdulkabir/CropDoc-Nigeria

Offline multimodal AI that diagnoses crop diseases from photos # 🌿 CropDoc Nigeria ### Offline Multimodal AI Crop Disease Detector — Powered by Gemma 4 > **Build with Gemma: GDGoc LASU Hackathon 2026** > Track: Agriculture & Food Security --- ## The Problem Nigeria's **37 million smallholder farmers** lose 30–40% of every harvest to crop diseases they cannot identify. No internet. No nearby expert. Wrong treatment makes it worse. ## The Solution **CropDoc Nigeria** — point your phone at any crop, get an instant AI diagnosis in **English, Yoruba, or Nigerian Pidgin** — completely offline. ``` 📸 Upload crop photo ↓ 🤖 Gemma 4 Vision analyzes image ↓ 🔬 Structured JSON diagnosis returned ↓ 🌍 Report in English / Yoruba / Pidgin ↓ 💊 Treatment plan with local products ``` --- ## How Gemma 4 Powers Everything | Gemma 4 Capability | How CropDoc Uses It | |---|---| | **Multimodal Vision** | Reads raw crop photo pixels → identifies disease from visual symptoms | | **Structured Output** | Returns clean JSON diagnosis via prompt engineering | | **Multilingual Gen.** | Generates Yoruba + Pidgin responses natively — zero hardcoding | | **On-device Inference** | Runs on local GPU — no internet after model load | > Remove Gemma 4 → app produces nothing. Gemma **is** the intelligence layer. --- ## Crops & Diseases Covered | Crop | Diseases | |---|---| | 🌿 Cassava (Igi Kara) | Mosaic Disease (CMD), Brown Streak (CBSD), Bacterial Blight | | 🍅 Tomato (Tomati) | Early Blight, Leaf Curl Virus, Late Blight | | 🌽 Maize (Agbado) | Streak Virus (MSV), Northern Corn Leaf Blight | | 🍠 Yam (Isu) | Anthracnose | | 🫑 Pepper, Cowpea, Plantain | Common fungal + viral diseases | --- ## Quickstart ### Prerequisites - Kaggle account with **T4 GPU** enabled - Gemma 4 model attached from Kaggle Models hub - Internet enabled (for first-time model download) ### 1. Clone this repo ```bash git clone github.com cd cropdoc-nigeria ``` ### 2. Install dependencies ```bash pip install git+github.com …