Outdoor Accessibility Object Detection Dataset (OAODD)

---

Overview

The Outdoor Accessibility Object Detection Dataset (OAODD) is a manually annotated object detection dataset consisting of diverse outdoor urban scenes. It contains high-quality bounding box annotations for commonly encountered objects and environmental elements in public spaces.

The dataset was created to support research and development in computer vision and object detection, providing a benchmark for training and evaluating deep learning models under real-world outdoor conditions.

---

Dataset Statistics

Property	Value
Task	Object Detection
Images	2,988
Annotations	9,089
Classes	9
Average Objects per Image	3.0
Missing Annotations	0
Median Resolution	960 × 1280
Annotation Type	Bounding Boxes

---

Classes

The dataset contains annotations for the following object categories:

• Car
• Chair
• Hole
• Obstacle
• Person
• Sidewalk
• Stairs
• Trash
• Tree

---

Dataset Structure

OAODD/

├── train/
│   ├── images/
│   └── labels/
│
├── valid/
│   ├── images/
│   └── labels/
│
├── test/
│   ├── images/
│   └── labels/
│
├── preview/
├── data.yaml
├── LICENSE
└── README.md

---

Dataset Preview

Example images from the dataset:

---

Annotation Format

The dataset is provided in YOLO Object Detection format.

Each image has a corresponding label file containing:

class x_center y_center width height

where all coordinates are normalized.

---

Intended Use

This dataset is intended for research and educational purposes, including:

• Object Detection
• Computer Vision Research
• Deep Learning
• Urban Scene Understanding
• AI Benchmarking
• Academic Projects

---

Citation

If you use this dataset in your research, please cite:

@dataset{boarat2026oaodd,
  author = {Dana Bo'arat},
  title = {Outdoor Accessibility Object Detection Dataset (OAODD)},
  year = {2026},
  version = {1.0}
}

---

License

This dataset is distributed under the Creative Commons Attribution 4.0 International (CC BY 4.0) License.

For more information, see the LICENSE file.

---

Contact

Dana Bo'arat

📧 danaboayrat@gmail.com