Abstract

We introduce UrbanSyn, a photorealistic dataset acquired throughsemi-procedurally generated synthetic urban driving scenarios. Developed usinghigh-quality geometry and materials, UrbanSyn provides pixel-level groundtruth, including depth, semantic segmentation, and instance segmentation withobject bounding boxes and occlusion degree. It complements GTAV and Synscapesdatasets to form what we coin as the 'Three Musketeers'. We demonstrate thevalue of the Three Musketeers in unsupervised domain adaptation for imagesemantic segmentation. Results on real-world datasets, Cityscapes, MapillaryVistas, and BDD100K, establish new benchmarks, largely attributed to UrbanSyn.We make UrbanSyn openly and freely accessible (www.urbansyn.org).