We present Fishyscapes, the first public benchmark for anomaly detection in a real-world task of semantic segmentation for urban driving. Existing uncertainty estimates have mostly been evaluated on simple tasks, and it is unclear whether these methods generalize to more complex scenarios. Yet, the ability to estimate uncertainty and detect failure is key for safety-critical applications like autonomous driving. Our models perform competitively with respect to the state of the art despite producing predictions with only one forward pass.ĭeep learning has enabled impressive progress in the accuracy of semantic segmentation. We present image classification experiments on CIFAR-10, as well as semantic segmentation experiments on three existing datasets (StreetHazards, WD-Pascal, Fishyscapes Lost & Found), and one contributed dataset. We show that this approach can be adapted for simultaneous semantic segmentation and dense outlier detection. We obtain the synthetic outliers by sampling an RNVP model which is jointly trained to generate datapoints at the border of the training distribution.

We consider an outlier detection approach based on discriminative training with jointly learned synthetic outliers. Consequently, this vulnerability represents an important research challenge. Interestingly, both discriminative and generative models appear to be equally affected. This gives rise to confident incorrect predictions which could lead to devastating consequences in many important application fields such as healthcare and autonomous driving. Today's deep models are often unable to detect inputs which do not belong to the training distribution.

Dense open set contains a subshift