🥔 Potato Disease Classification - Deep Learning Project

A deep learning web application built with TensorFlow, FastAPI, and LangChain that classifies potato leaf diseases and provides an AI-powered agricultural chatbot for farmers and researchers.

The app predicts whether a potato leaf is:

• Early Blight 🍂
• Late Blight 🧫
• Healthy ✅

---

🚀 Features

• 📷 Image Upload & Real-Time Disease Prediction
• 💬 RAG-based Agricultural Chatbot integrated using LangChain
• 🎨 Clean and responsive frontend (HTML, CSS, JS)
• 🌆 Background image with blur effect
• 🧠 Convolutional Neural Network (CNN) for disease classification
• ⚡ FastAPI Backend for prediction and chatbot APIs
• 🔒 .env file support for secret keys (e.g., Hugging Face token)

---

🧠 Model Overview

• Input Shape: 256x256x3
• Architecture:
  • 6 × Conv2D layers (ReLU activation)
  • MaxPooling for downsampling
  • Dense layers for classification
  • Softmax output layer
• Model File: potatoes.h5
• Framework: TensorFlow / Keras

---

🤖 RAG-Based Chatbot Integration

Alongside disease prediction, this project includes a Retrieval-Augmented Generation (RAG) chatbot that answers agricultural queries related to potato diseases, treatment methods, and farming practices.

🧩 Chatbot Pipeline

1. Knowledge Base Creation

   • Collected domain-specific text documents about potato diseases and farming techniques.
   • Loaded using TextLoader from LangChain.

2. Text Processing

   • Split into smaller chunks using RecursiveCharacterTextSplitter for efficient retrieval.

3. Vector Store

   • Created embeddings using:

     model_name = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"

   • Stored embeddings in FAISS (Facebook AI Similarity Search) for high-speed retrieval.

4. RAG Chain

   • Combined retriever + Hugging Face language model for contextual question answering.
   • Deployed via FastAPI endpoints alongside the disease classification API.

---

🎥 Demo Video

Watch the full project demo on YouTube: 👉 https://youtu.be/ZknDwnZHyRk

⚙️ Installation & Setup

✅ 1. Clone the repo

bash git clone https://github.com/aliahmad552/potato_disease_recognition.git cd Potato-Disease-Classification

✅ 2. Install dependencies

bash Copy Edit pip install -r requirements.txt

✅ 3. Run the Flask app

bash Copy Edit python app.py Now visit: http://127.0.0.1:5000

📂 Dataset

Custom dataset with three classes (Early Blight, Late Blight, Healthy)

Train/Validation/Test split handled using ImageDataGenerator

Image size: 256x256, normalized between 0-1

💻 Author

Ali Ahmad BS Software Engineering – The Islamia University of Bahawalpur GitHub: aliahmad552