Dataset Card for Planespotting Dataset

Dataset Summary

This dataset contains a collection of aircraft images for object detection purposes, focused specifically on amateur photography taken with ordinary smartphones or low-quality airport cameras. The dataset includes approximately 10,000 images with machine-generated annotations in YOLO format for identifying aircraft in various contexts. The primary goal is to detect aircraft on approach, often appearing as small dots in the sky, followed by the ability to auto-detect and auto-photograph in good quality. The dataset is balanced with roughly 50% low-quality images (showing very distant aircraft) and 50% high-quality images (showing aircraft at understandably close range). Images are organized into training and validation splits, with bounding box annotations that specify the location of aircraft within each image. This dataset is suitable for training and evaluating computer vision models for aircraft detection and related applications in aerial imagery analysis, aviation monitoring, and surveillance systems.

Dataset Structure

Data Files

The dataset is organized as follows:

• images/train/: Contains 9,885 training images
• images/val/: Contains 93 validation images
• labels/train/: Contains YOLO format annotation files for training images
• labels/val/: Contains YOLO format annotation files for validation images
• dataset.yaml: Configuration file specifying dataset paths and class names

Data Fields

Each image in the dataset is accompanied by a corresponding text file in YOLO format with the following structure:

• Each line represents one object annotation
• Format: class_id center_x center_y width height
  • class_id: Integer representing the object class (0 for aircraft)
  • center_x, center_y: Normalized coordinates (0-1) of the bounding box center
  • width, height: Normalized width and height (0-1) of the bounding box

Data Splits

The dataset contains two splits:

Split	Number of Examples
train	9,885
val	93

Total dataset size: 9,978 entries

Class Distribution

The dataset contains a single class:

• Class 0: aircraft

Additional Information

Annotation Process

Annotations were machine-generated, creating bounding boxes around aircraft in the images. The annotation format follows the YOLO standard, which uses normalized coordinates:

• First value: class ID (0 for aircraft)
• Second and third values: x and y coordinates of the bounding box center (normalized to 0-1)
• Fourth and fifth values: width and height of the bounding box (normalized to 0-1)

Example annotation: 0 0.497070 0.598633 0.993247 0.528925

License Information

The dataset is provided under the Creative Commons Zero (CC0) license, which effectively places it in the public domain. This means:

• No attribution is required for use
• The dataset can be used for both personal and commercial purposes
• Contents can be modified and redistributed without restrictions

For detailed information about the CC0 license, please refer to: https://creativecommons.org/publicdomain/zero/1.0/