bbarafat/disease-detection

MCF-MDS Africa day 2026 project. This project develops a web-based chatbot that allows farmers to upload images of pest/disease riddled crops, and the chatbot classifies the issue and provides suggestions for treatment --- title: Disease Detection emoji: 🌱 colorFrom: green colorTo: yellow sdk: streamlit sdk_version: 1.40.0 app_file: app.py pinned: false --- # Disease Detection MCF-MDS Africa day 2026 project. This project develops a web-based chatbot that allows farmers to upload images of pest/disease riddled crops, and the chatbot classifies the issue and provides suggestions for treatment We will use a pretrained model like EfficientNet, MobileNet, and Vision Transformers (ViT) to enhance accuracy \# Disease Detection Chatbot MCF-MDS Africa Day 2026 project: a web-based crop disease assistant for farmers. Farmers upload a crop leaf image, the app predicts the likely crop disease using a fine-tuned computer vision classifier, then an LLM-powered assistant helps explain the result and suggests practical treatment and prevention steps. > Important: this app is an educational decision-support tool, not a replacement for a trained agronomist or local extension officer. ## Core idea 1. Train an image classifier on the PlantVillage dataset. 2. Save the trained model and class labels. 3. Serve predictions through a FastAPI backend. 4. Use an LLM provider abstraction so we can start with a free/local Hugging Face model and later swap in OpenAI, Anthropic, Gemini, or another provider. 5. Build a simple Streamlit frontend where a farmer can upload an image and ask follow-up questions. ## Target stack - Python 3.10+ - PyTorch + torchvision for the crop disease classifier - FastAPI for the backend API - Streamlit for the web UI - Hugging Face Transformers for a free/local LLM fallback - Optional future LLM providers: OpenAI, Anthropic, Gemini, Hugging Face Inference Providers ## Proposed repository structure ``` text disease-detection/ ├── app/ │ ├── api/ │ │ └── main.py │ ├── core/ │ │ ├── config.py │ │ ├── disease_knowledge.py │ │ ├── image_model.py │ │ └── llm.py │ └── streamlit_app.py ├── data/ │ ├── raw/ │ └── processed/ ├── mo …