OoDIS: Anomaly Instance Segmentation Benchmark
Alexey Nekrasov1 • Rui Zhou1,3 • Miriam Ackermann2 • Alexander Hermans1 • Bastian Leibe1 • Matthias Rottmann4
1 RWTH Aachen University (Germany)
2 Ruhr University Bochum (Germany)
3 Beijing Institute of Technology (China)
4 Osnabrück University (Germany)
International Conference on Robotics and Automation (ICRA), 2025
Anomaly Instance Segmentation is a task that aims to find objects that are present only at inference time and unknown during training. A typical anomaly is a deer or a cardboard box in the middle of the road. In the benchmark, we extend the labels of well-known benchmarks such as SegemntMeIfYouCan and FishyScapes Lost and Found to instance segmentation. We combine two benchmarks into a unified benchmark and evaluate the most common metrics instance metrics of Average Precision.
We extend labels of 2 existing benchmarks for Instance Segmentation:
• SegmentMeIfYouCan, containing two datasets: RoadAnomaly21 and RoadObstacle21
• Fishyscapes: Lost & Found
The benchmark page is available here.
Resources & Links
Acknowledgments
M.A. and M.R. acknowledge support by the BMBTR, grant 01IS22069.
Citation
@inproceedings{nekrasov2025oodis,
title={OoDIS: Anomaly Instance Segmentation and Detection Benchmark},
author={Nekrasov, Alexey and Zhou, Rui and Ackermann, Miriam and Hermans, Alexander and Leibe, Bastian and Rottmann, Matthias},
booktitle={2025 IEEE International Conference on Robotics and Automation (ICRA)},
pages={2764--2771},
year={2025},
organization={IEEE}
}
Contact
Have questions or want to collaborate? Reach out: