Nao Robot Soccer Image Segmentation Dataset

This repository contains the public website and paper source for a large RoboCup robot soccer image dataset collected from Nao V6 humanoid robots.

• Website: https://hulks.de/nao-image-segmentation-dataset/
• Dataset download: https://nextcloud.hulks.dev/s/RfEkfeKTk6H6aJX
• Contact: hulks@tuhh.de

Dataset

The dataset contains 1,864,394 images captured by Nao V6 robots across teams, venues, and years. Each game is manually categorized by environmental conditions that affect robot soccer vision.

• Light source: sun, artificial, and mixed lighting
• Shadows and reflections: none, reflections, shadows, or both
• Line conditions: taped or spray-painted field markings
• Field conditions: consistent or inconsistent field color

A representative subset of 601 images includes pixel-wise segmentation masks for field, lines, ball, robots, goal, others, and uncertain pixels.

Repository Layout

• config.toml: Zola site configuration
• templates/: Zola HTML templates
• sass/: site styles
• static/: static website assets copied into the generated site
• paper/: LNCS/LNAI paper source, figures, and local build artifacts

Local Website Preview

Install Zola 0.22 or newer, then run:

zola serve

Paper Build

The manuscript sources are in paper/. To rebuild the PDF locally:

cd paper
latexmk paper.tex

Citation

@misc{goettsch2026naodataset,
  title  = {A Large Image Dataset for Robot Soccer in Diverse Competition Environments},
  author = {G\"ottsch, Franziska-Sophie and Schmidt, Maximilian},
  year   = {2026},
  note   = {RoboCup International Symposium Open Platforms and Tools submission}
}