UAV-and-Bird-Identification

This repository consists of the codebase for the UAV and Bird identification model.

Table of Contents

• Intrduction
• Dataset
• File Strcuture
• Screenshots
• Results
• Installation

Introduction

The model here is trained to identify birds and UAVs (Drones) for safety purposes. The model proposed an accuracy above 90% in all classes and for Birds and UAVs separately. The ultimate goal of developing this model was to run it in real-time with lower inference time and better accuracy when compared to other lightweight models

Dataset

This is a custom dataset that integrates multiple datasets for more scenarios. The following are the data instances for each defect.
Images Instances
Birds 3581 22620
UAV 3334 3571

File Structure

UAVBirds/
├── merged/

│ └── Images/
│ │ └── Test/

│ │ ├── Train/

│ │ └── Valid/

│ ├── Labels/

│ │ └── Test/

│ │ ├── Train/

│ │ └── Valid/

│ └── data,yaml

├── Weights #best and last .pt files
└── README.md

Screenshots

The following image is based on the model training. Training summary;

Results

Training results;

Installation

# Clone the repository
git clone https://github.com/ultralytics/ultralytics
cd ultralytics

# Install dependencies (choose one)
pip install -r requirements.txt

## Download the model
https://github.com/tikiya99/UAV-and-Bird-Identification.git