Leaderboard

Evaluation Metrics

The OoD Data flag indicates whether the method uses out-of-distribution data during training or fine-tuning.
We consider the Cityscapes dataset as the only in-distribution training data.
Any additional datasets are categorized as OoD data.
We encourage readers to also check the original papers of the methods to double check the use of OoD data.

For a more detailed explanation of the metrics and OoD data, we refer to our paper.



Anomaly Track

Obstacle Track

LostAndFound NoKnown



For a description of the test datasets, we refer to our paper.