Real-Time Wild Animal Detection and Alert System Using Machine Learning

Project Overview

Wild animal intrusion into human settlements is a serious issue in forest-covered regions like Wayanad, Kerala, leading to loss of human life, crop damage, and economic loss. Existing solutions such as electric fencing are often ineffective and unsafe.

This project implements a real-time automated system that detects wild animals using computer vision and deep learning, triggers species-specific deterrent responses, and sends instant alerts to concerned authorities.

Objectives

• Detect wild animals (Elephant, Tiger, Wild Boar) from video input in real-time
• Trigger species-specific sound and light deterrents upon detection
• Send automated alerts via WhatsApp/SMS to authorized personnel
• Reduce human-animal conflict using a non-harmful, automated system

Technical Implementation

• Programming Language: Python 3.8+
• Deep Learning Model: YOLOv8 (Ultralytics implementation)
• Machine Learning Approach: Supervised Learning with Transfer Learning
• Detection Framework: Real-time object detection using PyTorch backend

System Architecture

1. Video Input Capture: OpenCV processes live video feed
2. Object Detection: YOLOv8 model processes frames for animal detection
3. Response Trigger:
   • Species-specific deterrent activation (sound/light)
   • Alert notification via Twilio API
4. Alert Management: One-time alert per video to prevent notification spam

Animal Classes Detected

The system is trained to detect three primary wildlife species:

• Elephant
• Tiger
• Wild Boar

Dataset

• Source: Roboflow Public Dataset (CC BY 4.0 License)
• Total Images: Approximately 1,800 labeled images
• Split Ratio: 70% Training, 20% Validation, 10% Testing
• Annotation Format: YOLOv8 compatible bounding boxes

Deterrent System

Animal	Sound Response	Light Response
Elephant	Drum / Bee sounds	Red + White strobe
Tiger	Siren / Dog bark	White strobe
Wild Boar	Alarm sound	White + Green strobe

Deterrent responses are based on documented animal behavior studies and are non-harmful.

Alert System

• Platform: Twilio API for WhatsApp/SMS delivery
• Behavior: Single alert per video session to avoid notification overload
• Content: Includes detected animal name and confidence score
• Recipients: Configurable list of authorized contacts

Model Performance

• Precision: ~89%
• Recall: ~85%
• Mean Average Precision (mAP): ~88%
• Training: 11-100 epochs using transfer learning from COCO pretrained weights

Technical Features

• Real-time Processing: 15-30 FPS depending on hardware
• Low-light Handling: Brightness enhancement and histogram equalization
• Modular Design: Separate modules for detection, deterrents, and alerts
• Configuration Management: Environment variables for API keys and settings

Installation & Setup

Prerequisites

• Python 3.8 or higher
• Git

Installation Steps

1. Clone the repository:

   git clone https://github.com/your-username/wild-animal-detection.git
   cd wild-animal-detection

Install dependencies:

bash

pip install -r requirements.txt

Configure environment variables:

Copy .env.example to .env

Add your Twilio API credentials:

text

TWILIO_ACCOUNT_SID=your_account_sid

TWILIO_AUTH_TOKEN=your_auth_token

TWILIO_FROM=your_twilio_number

TWILIO_TO=target_phone_number

Download the dataset (if training):

Visit: https://universe.roboflow.com/hbz-syqzm/tiger-elephant-boar

Download in YOLOv8 format

Extract to data/roboflow_dataset/

Running the System

bash

python src/detection.py

Project Structure

text

wild-animal-detection/

├── src/ # Source code

│ ├── detection.py # Main detection script

│ ├── alert_system.py # Twilio alert integration

│ └── deterrents.py # Sound/light deterrent controls

├── data/ # Dataset documentation

│ └── README.md # Dataset instructions

├── models/ # Trained model weights

├── requirements.txt # Python dependencies

├── .env.example # Environment template

├── .gitignore # Git ignore rules

└── README.md # This file

Limitations

Night detection accuracy requires additional training on infrared/thermal datasets

Limited to three animal classes in current implementation

Dependent on camera quality and environmental conditions

Twilio free tier has message limits

Future Enhancements

Integration with thermal/IR cameras for night detection

Mobile application dashboard for real-time monitoring

Hardware deployment using Raspberry Pi with connected deterrent devices

Multi-camera support with cloud-based processing

GPS-based location tagging in alerts

Academic Context

This project was developed as part of the Master of Computer Applications (MCA) program, demonstrating practical application of machine learning and computer vision concepts to real-world environmental challenges.

Testing & Validation

The system has been validated on sample wildlife videos with the following outcomes:

Correct animal detection and species identification

Appropriate deterrent activation based on detected species

Successful alert delivery via configured channels

No false positives for non-target animals/objects

License

This project is available for academic and educational purposes. The dataset used is licensed under CC BY 4.0.

Contact

For questions or collaboration inquiries, please contact:

Name: Shaheer Ali S B

Email: alishaheer272002@gmail.com

GitHub: https://github.com/shaheer-ali-sb

Last Updated: December 2025