🌿 Plant Disease Chatbot

An AI-powered conversational chatbot for diagnosing plant diseases using computer vision (CNN) and natural language processing (BERT).

✨ Features

• 🤖 Conversational AI: Interactive chatbot that asks clarifying questions
• 📸 Image Diagnosis: Upload leaf photos for CNN-based disease detection
• 💬 Text Diagnosis: Describe symptoms for BERT-based analysis
• 🎯 Pattern-Based Detection: Recognizes characteristic disease patterns (e.g., Early Blight's target spots)
• 💊 Treatment Recommendations: Detailed treatment and prevention advice for 35+ diseases
• 🌍 Multi-Crop Support: Covers tomatoes, potatoes, apples, grapes, corn, peppers, and more

🚀 Quick Start

Prerequisites

• Python 3.8+
• pip

Installation

1. Clone the repository:

git clone <your-repo-url>
cd final

2. Create virtual environment:

python -m venv .venv
.venv\Scripts\activate  # Windows
# or
source .venv/bin/activate  # Linux/Mac

3. Install dependencies:

pip install -r requirements.txt

4. Download model files (not included in Git due to size):

   • bert_plant_chatbot_model.pt (438 MB)
   • cnn_model_final.h5 or final_saved_model/ (27-574 MB)
   • class_names.pkl
   • label_mapping.json

   Note: Contact the repository owner for model files or train your own using Training_code.ipynb

5. Run the app:

streamlit run app.py

6. Open in browser: http://localhost:8501

📁 Project Structure

final/
├── app.py                      # Main Streamlit application
├── chatbot/                    # Chatbot package
│   ├── chatbot_manager.py      # Conversational AI logic
│   ├── conversation_templates.py # Response templates
│   └── followup_templates.py   # Follow-up templates
├── knowledge_base/             # Knowledge base package
│   ├── disease_patterns.py     # Pattern-based disease detection
│   ├── treatments.py           # Treatment database (35 diseases)
│   └── detailed_treatments.py  # Detailed disease info
├── resources/                  # Static resources
│   ├── class_names.pkl         # CNN labels (not in Git)
│   └── label_mapping.json      # BERT labels (not in Git)
├── models/                     # Model files
│   ├── bert_plant_chatbot_model.pt # BERT model (not in Git)
│   └── final_saved_model/      # CNN model (not in Git)
├── requirements.txt            # Python dependencies
├── README.md                   # This file
└── .gitignore                  # Git ignore rules

🎯 Usage

1. Image Diagnosis

• Upload a clear photo of affected plant leaves
• Get instant CNN-based disease prediction
• View confidence score and treatment recommendations

2. Text Diagnosis

• Describe symptoms in natural language
• Get BERT-based analysis with top 3 predictions
• See detailed treatment plans

3. Chat Diagnosis (Conversational)

• Start a conversation with the AI assistant
• Answer clarifying questions about symptoms
• Receive accurate diagnosis with high confidence
• Get personalized treatment recommendations

Example Conversation:

🤖 Bot: Hello! Describe the symptoms you're seeing.

👤 User: My tomato plant has brown spots with rings on the leaves

🤖 Bot: Are the spots circular with concentric rings, like a target?

👤 User: Yes, exactly like a target pattern

🤖 Bot: I'm 70% confident this is Tomato Early Blight.
       
       Treatment: Remove infected leaves, apply fungicides...

🧠 How It Works

Dual Detection System

1. Pattern-Based Detection (Primary):

   • Recognizes characteristic symptom patterns
   • High confidence (70-95%) for specific diseases
   • Examples: Early Blight (target spots), Powdery Mildew (white powder)

2. BERT Model (Fallback):

   • Natural language understanding
   • Analyzes accumulated symptom descriptions
   • Works for diseases without specific patterns

Supported Diseases (35+)

Crops: Tomato (12 conditions), Potato (3), Apple (3), Corn (3), Grape (2), Pepper (3), and more

Disease Types: Fungal, bacterial, viral, pest-related, and healthy baselines

🛠️ Technical Stack

• Frontend: Streamlit
• ML Models:
  • TensorFlow/Keras (CNN for images)
  • PyTorch + Transformers (BERT for text)
• Image Processing: PIL
• Data: NumPy, Pickle, JSON

⚠️ Important Notes

Model Files Not Included

Due to GitHub file size limits, the following large files are not included in this repository:

• bert_plant_chatbot_model.pt (438 MB)
• final_saved_model/ or cnn_model_final.h5 (27-574 MB)
• newmodel.h5 (574 MB)

To get the models:

1. Contact the repository owner
2. Or train your own using Training_code.ipynb
3. Or use Git LFS for large files

Disclaimer

This tool provides general guidance and is not a replacement for professional diagnosis. For critical plant health issues, consult certified agricultural professionals.

📊 Performance

• Pattern Detection: 70-95% confidence for characteristic diseases
• BERT Model: 10-20% confidence (requires improvement)
• CNN Model: Varies by disease
• Inference Speed: 1-3 seconds on CPU

🔮 Future Enhancements

• Improve BERT model training
• Add more disease patterns
• Mobile app version
• Multi-language support
• User accounts and history
• Geographic disease tracking
• Integration with weather data

🤝 Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues.

📄 License

[Add your license here]

👥 Authors

[Add your name/team here]

🙏 Acknowledgments

• PlantDoc and PlantVillage datasets
• BERT model from Hugging Face
• Streamlit for the amazing framework

---

Built with ❤️ for healthier plants 🌱