ADR Sign Sample Dataset for Drones See Signs

DMP DOI: https://doi.org/10.5281/zenodo.17769823

Context and Methodology

This dataset was created as part of the Drones See Signs project, an in-progress research project focused on detecting ADR (hazard) signs during drone flight using YOLO-based object detection methods. The purpose of this dataset is to provide an example of the type of training data required for object detection.
The dataset consists of a placeholder ADR sign image and a YOLO-format annotation file. These files illustrate the structure, metadata, and documentation practices expected in a real dataset, which will be generated later in the research.

Technical Details

The dataset is organized into two main folders:

• /images/ — contains hypothetical ADR sign images in PNG or JPEG format.

• /labels/ — contains hypothetical YOLO annotations in .txt format, where each file corresponds to an image of the same name.

A standard YOLO naming convention is used:
image_00X.png
image_00X.txt
Each annotation file contains one or more lines representing object class and normalized bounding-box coordinates.

No specialized software is required to view the images. The annotations can be opened with any text editor. For training experiments or replication, the dataset is compatible with any YOLO model.

Further Details

This dataset is intentionally minimal and serves as an sample for the Drones See Signs project. For further clarification, it is mentioned in the DMP of the project, that a dataset R1 from Roboflow universe is being re-used. It should be noted that this dataset therefore serves as an extensive custom dataset, for those who wishes to train the model on their own dataset for their own targeted object-of-interest.