SulemanaAbu/CattleCare-AI

AI-Based Cattle Disease Detection System for Ghanaian Farmers and Traders Final Year Project — University of Ghana, Department of Computer Science (2025/2026) # 🐄 CattleCare AI **AI-Based Cattle Disease Detection System for Ghanaian Farmers and Traders** Final Year Project — University of Ghana, Department of Computer Science (2025/2026) CattleCare AI helps smallholder farmers and cattle traders get an early, AI-assisted read on three common conditions — **Foot-and-Mouth Disease (FMD)**, **Lumpy Skin Disease (LSD)**, and **Healthy** — from a photo, a description of observed symptoms, or both. It is built around a Gemma 3 vision-language model fine-tuned specifically on Ghanaian cattle disease presentations, served through a Streamlit interface. > ⚠️ **Disclaimer:** This is a decision-support tool, not a replacement for veterinary diagnosis. Always consult a qualified veterinarian for confirmation and treatment. --- ## Features - **Three ways to get a diagnosis:** - 📷 Upload a photo of the animal - 📝 Describe symptoms (checkboxes + free text) - 🔗 Both together, for the most accurate result - **Input validation before diagnosis** — a lightweight classification layer checks that an uploaded photo actually shows cattle and that a symptom description is real and observable, rather than letting the model guess at chitchat, junk input, or irrelevant images. - **Transparent source attribution** — the UI shows exactly which inputs (image / symptom description) were actually used for a given diagnosis. - **Symmetric fallback logic** — a bad photo will not derail a real symptom description, and vice versa; the app automatically falls back to whichever input is actually usable. - **Model comparison mode** — base model vs. fine-tuned model, side by side, on the same input. - **Disease reference guide** — a built-in tab covering symptoms, spread, and recommended actions for each condition. - **Emergency contacts** — direct numbers for the Ghana Veterinary Service and Animal Research Institute surfaced alongside every diagnosis. --- ## How It Works ``` ┌─────────────────────┐ Photo + Symptoms │ Input Validation │ gpt-4o …