This repository contains a dataset of 60 real images and 100 synthetic images, designed to support research in vision-based damage risk estimation. The images are displayed below.
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
 |  |  |  |  |
Further details can be found in the following publication:
title={Damage Risk Quantification for Robot Collisions Using Vision-Language Models}, author = {Jonas Kiemel and Erhan \"Oztop and Tamim Asfour},
publisher = {German Robotics Conference (GRC)},
year = {2026}