Our German traffic dataset contains more than 300,000 images with more than 2 million individual labels in 280 classes. The class system is based on the German road traffic regulation (StVo) and covers traffic signs and lights. Spatial filters applied during the collection guarantee a minimum distance between individual images. The dataset contains city, villages, country roads, federal roads and highway scenes in all weather conditions, daytimes and time of year. It includes a variation of corner cases like extreme light, snow and rain to validate Deep Learning algorithms for traffic scene understanding. As the dataset is continuously extended we provide minimum numbers for some properties.

Resolution: 1920×1200 pixels
Camera: Basler Ace acA1920-40uc
Sensor: Sony IMX249
Lens: Tamron: M1112FM08
Format: JPEG (Quality Level 95-100%)
Size: 220 GB+
Classes: 280+
Images: 300.000+
Minimum distance between individual images: 4m

Further documentation:

Download class statistics

Download taxonomy

Download visual samples

Download test sample