ACDC is a new large-​scale driving dataset for training and testing semantic segmentation algorithms on adverse visual conditions, such as fog, nighttime, rain, and snow. The dataset and associated benchmarks are published at ICCV 2021 and are now publicly available at

The dataset consists of 4006 images which are distributed evenly between four common adverse conditions: fog, nighttime, rain, and snow. For each image, we provide a high-​quality pixel-​level semantic annotation, a corresponding image of the same scene taken under normal conditions, and a binary mask that distinguishes between intra-​image regions of clear and uncertain semantic content. At, we also provide a benchmark suite and evaluation server for the two tasks supported by ACDC: standard semantic segmentation and uncertainty-​aware semantic segmentation. ACDC serves as a test bed both for supervised semantic segmentation and unsupervised domain adaptation, especially in the normal-​to-adverse adaptation setting. Our associated paper is available on arXiv at