This project is an open vision initiative built using YOLOv8, a state-of-the-art object detection model developed by Ultralytics. It presents a complete pipeline for detecting animals from CCTV footage — suitable for both real-time and static image analysis.

The system is designed to be modular, scalable, and easy to deploy, and is particularly useful for:

• 🐄 Farm and livestock monitoring
• 🐘 Wildlife conservation and animal tracking
• 🚧 Roadside animal presence alerts
• 📷 General CCTV-based animal surveillance

This repository includes:

• ✅ Custom dataset loading and analysis
• ✅ Augmentation and data quality checks
• ✅ YOLOv8 model training and evaluation
• ✅ Inference interface for single images
• ✅ Example visualization of results

Source: Private Dataset

Classes:

['Bird', 'Cats', 'Cow', 'Deer', 'Dog', 'Elephant', 'Giraffe', 'Person', 'Pig', 'Sheep']

Folder Structure:

└── train
      ├── images
      └── labels
└── valid
      ├── images
      └── labels
└── test
      ├── images
      └── labels

Label Format (YOLO):

<class_id> <x_center> <y_center> <width> <height>   # All values normalized (0–1)

• Upload the .ipynb notebook to Kaggle or run locally with Jupyter.

pip install ultralytics albumentations opencv-python matplotlib seaborn

• Dataset structure check
• Class distribution visualization
• Sample image bounding box rendering
• Augmentation and quality validation

• Loads pretrained yolov8s.pt
• Trains for 25 epochs, imgsz=640, batch=16
• Saves outputs to runs/detect/animal-detection-yolov8/

Generates plots automatically:

• 📈 F1 Curve
• 🎯 Precision-Recall Curve
• ✅ Precision and Recall

• Run inference on a static image
• Output is visualized directly using matplotlib
• Extendable to webcam/video input

Sample YOLOv8 output with bounding boxes on animal classes:

📷 Image: valid/images/sample.jpg
→ Detected: Cow, Dog, Sheep with confidence > 0.5
→ Rendered bounding boxes in real-time

YOLOv8 provides performance metrics during and after training, including:

• mAP@0.5, mAP@0.5:0.95
• Precision, Recall, and F1 score
• Loss curves for training and validation

All metrics are logged automatically and plots are available under:

runs/detect/animal-detection-yolov8/

This project uses Python 3.8+ and runs in environments like Kaggle, Colab, or Jupyter.

pip install ultralytics albumentations opencv-python matplotlib seaborn

The trained YOLOv8 weights aren’t published here please email champion19007@gmail.com to request access.

Sai Yashwant Reddy Panthy
AI/ML Developer
📧 Email: champion19007@gmail.com

Feel free to reach out for collaboration or inquiries.