UAV Detection Dataset & YOLOv5 Training Guide

Overview

This repository contains the dataset and scripts needed for training a YOLOv5 model to detect UAVs (Unmanned Aerial Vehicles) from images captured at multiple angles. The dataset is specifically curated to provide diverse perspectives of UAVs, making it suitable for robust object detection models.

Dataset

Downloading the Dataset

The dataset can be downloaded from the following link:

Download link

Dataset Structure

The dataset is organized as follows:

dataset/
├── images/
│   ├── train/
│   ├── val/
│   └── test/
└── labels/
    ├── train/
    ├── val/
    └── test/

images/: Contains the image files for training, validation, and testing.
labels/: Contains the corresponding label files in YOLO format for each image.

Dataset Details

Total Images:1623
Image Resolution: 1920x1080 pixels
Classes: 1 (enemy)

Training YOLOv5

Prerequisites

Python 3.7+
YOLOv5 Repository
PyTorch 1.7+
Other dependencies as listed in the YOLOv5 repository.

Setup

Clone the YOLOv5 repository:

git clone https://github.com/ultralytics/yolov5.git
cd yolov5

Install the required dependencies:
```
pip install -r requirements.txt
```
Copy the dataset to the yolov5/dataset/ directory.

Training

To train the YOLOv5 model on the UAV dataset:

Update the data.yaml file in the yolov5/data/ directory to include the path to your dataset and the number of classes.

Example data.yaml:
```
train: ../dataset/images/train
val: ../dataset/images/val
nc: 1  # number of classes
names: ['enemy']  # class names
```
Start training using the following command:
```
python train.py --img 640 --batch 16 --epochs 50 --data ./data/data.yaml --weights yolov5s.pt --cache
```
- --img: Image size for training.
- --batch: Batch size.
- --epochs: Number of epochs.
- --data: Path to the data.yaml file.
- --weights: Pre-trained weights to start training from.
- --cache: Caches images for faster training.
Monitor the training process using the provided metrics and adjust parameters as needed.

Evaluation

After training, evaluate the model on the test dataset:

python val.py --data ./data/data.yaml --weights runs/train/exp/weights/best.pt --img 640

License

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

Acknowledgments

Ultralytics YOLOv5 for providing a robust object detection framework.
Contributors who helped create and curate the dataset.

shivamkainth/iloc