🦌 Poaching Detection using Deep Learning and Flask

A sophisticated web application for detecting poaching activities in wildlife images using deep learning. Built with Flask, TensorFlow, and a modern responsive UI.

Created by Society Internship Group

---

✨ Features

• 🎯 Drag-and-Drop Detection: Upload single images via intuitive drag-and-drop interface
• 📁 Batch Processing: Analyze entire folders of wildlife images at once
• 🤖 Deep Learning Model: Pre-trained CNN model (256x256 RGB input, binary classification)
• 📱 SMS Alerts: Automatic Twilio notifications when poachers are detected
• 🎨 Modern UI: Light theme with animated gradients, floating blobs, and smooth transitions
• ⚡ Real-time Results: Instant predictions with confidence scores and visual indicators
• 📊 Detailed Analytics: Probability percentages and classification confidence displayed

---

🚀 Quick Start

Prerequisites

• Python 3.13 (or compatible version)
• Windows/Linux/macOS
• 4GB RAM minimum (8GB recommended for model loading)

Installation

1. Clone or download this repository

   cd "C:\Users\DELL INSPIRON 3558\Desktop\SAI-II\Poaching-Detection-using-Deep-Learning-and-Flask"

2. Install dependencies

   py -m pip install -r requirement.txt

   This installs:

   • Flask 3.1.2
   • TensorFlow-CPU 2.20.0
   • Keras 3.12.0
   • OpenCV 4.12.0
   • Twilio (for SMS alerts)
   • NumPy, Matplotlib

3. Set up environment variables (Optional - for SMS alerts)

   Create environment variables for Twilio integration:

   $env:TWILIO_SID = "your_account_sid"
   $env:TWILIO_AUTH_TOKEN = "your_auth_token"
   $env:TWILIO_FROM = "+1234567890"
   $env:TWILIO_TO = "+0987654321"

4. Run the application

   py app.py

5. Access the web interface

   Open your browser and navigate to:

   http://127.0.0.1:5000
   

---

📖 Usage Guide

Method 1: Drag-and-Drop (Single Image)

1. Navigate to the "Quick Detection" section
2. Drag an image file (JPG, PNG, BMP, TIFF, WEBP) onto the drop zone
   • Or click the drop zone to select a file
3. The app will:
   • Display image preview with loading animation
   • Preprocess the image (BGR→RGB, resize 256x256, normalize 0-1)
   • Run prediction using the trained model
   • Show result with:
     • ✓ Green alert: No poacher detected
     • ⚠️ Red alert: Poacher detected
     • Confidence percentage
     • Progress bar visualization

Method 2: Folder Processing (Batch Detection)

1. Navigate to the "Batch Detection" section
2. Enter the absolute path to a folder containing wildlife images:

   C:\Users\YourName\Desktop\wildlife_images
   

3. Click "Run Detection"
4. The app will:
   • Process all images in the folder (JPG, JPEG, PNG, BMP, TIFF)
   • Display results for each image
   • Send SMS alert if poachers are detected (if Twilio configured)
   • Show summary of detections

---

🏗️ Project Structure

Poaching-Detection-using-Deep-Learning-and-Flask/
├── app.py                          # Main Flask application
├── requirement.txt                 # Python dependencies
├── README.md                       # This file
├── models/
│   └── poachingdetectionVER7.h5   # Pre-trained Keras model (256x256 RGB, binary)
├── static/
│   ├── script.js                  # Client-side JavaScript (drag-drop, animations)
│   └── css/
│       └── style.css              # Light theme styling, gradients, animations
├── templates/
│   ├── index.html                 # Main UI template (Bootstrap 5.3)
│   └── backup.html                # Backup template
└── videoconversion/
    ├── converter.py               # Video to frame extraction utility
    ├── copyofconverter.py         # Alternative converter
    └── removefile.py              # Cleanup utility

---

🔧 Technical Details

Model Architecture

• Input: 256x256 RGB images
• Type: Binary classification (Poacher vs No Poacher)
• Threshold: Probability > 0.5 = Poacher detected
• Loading: compile=False to avoid Keras 3.x reduction errors

Image Preprocessing Pipeline

1. Load image with OpenCV (cv2.imread)
2. Convert BGR → RGB (cv2.cvtColor)
3. Resize to 256x256 (cv2.resize)
4. Convert to float32 and normalize [0-1] (img.astype('float32') / 255.0)
5. Add batch dimension (np.expand_dims)
6. Predict with model

API Endpoints

GET /

• Renders main UI template
• Handles folder path form submissions (POST)

POST /predict-image

• Accepts: multipart/form-data with image field
• Validates: File type, dimensions, integrity
• Returns: JSON response

  {
    "is_poacher": true,
    "probability": 0.87,
    "confidence": 0.87,
    "message": "⚠️ Poacher detected!",
    "filename": "wildlife_cam_001.jpg"
  }

---

🎨 UI Features

Light Theme Design

• Color Scheme: Teal (#16a085) and Green (#20c997) accents
• Font: Poppins (Google Fonts)
• Icons: Bootstrap Icons 1.11
• Gradients: Soft blue-to-teal background with animated floating blobs

Animations

• Hero Section: Slide-in animations for title and subtitle
• Floating Icons: Continuous gentle floating motion
• Drop Zone: Hover effects with border color changes and scaling
• Results: Pop animation with scale transform
• Progress Bars: Smooth width transitions

Responsive Design

• Mobile-friendly layout with Bootstrap 5.3 grid
• Responsive font sizes and padding
• Touch-optimized drag-and-drop

---

🐛 Troubleshooting

Issue: pip is not recognized

Solution: Use py -m pip install -r requirement.txt

Issue: ModuleNotFoundError: No module named 'imghdr'

Solution: Already fixed in current version (unused import removed)

Issue: ValueError: Invalid value for argument reduction='auto'

Solution: Model loads with compile=False to bypass Keras 3.x issue

Issue: TwilioRestException: Authentication Error

Solution: Set environment variables or app works without Twilio (graceful fallback)

Issue: Wrong predictions / low accuracy

Checklist:

• ✓ Ensure images are clear and well-lit
• ✓ Model expects 256x256 RGB images
• ✓ Preprocessing must convert BGR→RGB (OpenCV loads BGR by default)
• ✓ Normalization to [0-1] range is critical
• ✓ Check if model is trained on similar wildlife camera images

Issue: Drag-and-drop not working

Checklist:

• ✓ Check browser console for JavaScript errors
• ✓ Ensure script.js is loaded (view page source)
• ✓ Try clicking the drop zone to select file manually
• ✓ Verify file extension is supported (JPG, PNG, BMP, TIFF, WEBP)

Issue: Model not loading

Checklist:

• ✓ Verify models/poachingdetectionVER7.h5 exists
• ✓ Check file permissions (read access required)
• ✓ Ensure TensorFlow and Keras are correctly installed
• ✓ Check terminal output for specific error messages

---

🚀 Advanced Configuration

Custom Model Path

Edit app.py line ~18:

MODEL_PATH = os.path.join(BASE_DIR, 'models', 'your_model_name.h5')

Change Detection Threshold

Edit app.py prediction logic (default: 0.5):

is_poacher = prob > 0.5  # Change to 0.6 for stricter detection

Enable Debug Mode

Edit app.py last line:

app.run(debug=True, host='0.0.0.0', port=5000)

Add Authentication

Install Flask-Login:

py -m pip install flask-login

Then add login routes and decorators to protect endpoints.

---

📊 Performance Optimization

Recommendations for Production

1. Use Gunicorn/Waitress instead of Flask development server:

   py -m pip install waitress
   waitress-serve --port=5000 app:app

2. Enable Model Caching: Already implemented via GLOBAL_MODEL

3. Add Redis for Session Management:

   py -m pip install flask-session redis

4. Implement Rate Limiting:

   py -m pip install flask-limiter

5. Use CDN for Static Files: Host Bootstrap, fonts, icons on CDN

---

🤝 Contributing

Contributions are welcome! Please follow these guidelines:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

---

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

🎓 Credits

Developed by: Society Internship Group

Technologies:

• Flask (Web Framework)
• TensorFlow/Keras (Deep Learning)
• OpenCV (Image Processing)
• Bootstrap 5 (UI Framework)
• Twilio (SMS Notifications)

---

📞 Support

For issues, questions, or contributions:

• Open an issue on GitHub
• Contact: Society Internship Group

---

🔮 Future Enhancements

• Real-time video stream analysis
• Multi-class detection (different animal species)
• Heat map visualization of detection regions
• User authentication and role-based access
• Historical analytics dashboard
• Export detection reports as PDF
• Integration with wildlife monitoring systems
• Mobile app (iOS/Android)
• Docker containerization
• Cloud deployment (AWS/Azure/GCP)

---

Made with ❤️ by Society Internship Group