> ML_DATASET // CITYSCAPES-DATASET_v1.0
Cityscapes: Semantic Understanding of Urban Street Scenes
Daimler AG / Max Planck Institute / TU Darmstadt (Cordts et al.) · Autonomous Driving & Urban Perception · 5,000 fine-annotated stereo images across 50 cities, 20,000 coarsely annotated frames
Autonomous Driving & Urban PerceptionCustom Research Non-Commercial License5,000 fine-annotated stereo images across 50 cities, 20,000 coarsely annotated framesregistration-required
Dataset Profile & Characteristics
Label Type:Pixel-level and instance-level semantic annotations across 30 visual classes
Languages:en
License Tier:non-commercial
Modalities:image, video
Intended Use
- Benchmarking urban semantic segmentation, instance segmentation, and panoptic driving perception
Prohibited / Discouraged Use
- Direct autonomous driving deployment without nighttime and adverse weather adaptation
Bias, Leakage & Privacy Risk Analysis
Privacy / Sensitive Data Risks:
Real-world German and European street driving scenes containing pedestrians and vehicle license plates.
Known Bias:
Central European daytime good-weather urban driving conditions.
Known Benchmark Leakage:
Video sequences from same driving route must be strictly kept in same split to prevent video correlation leak.