<p>In the field of autonomous robotics on public roads, accurate perception of the environment is crucial for safe and efficient operation. In particular, the training of algorithms for object detection and semantic segmentation requires datasets that can map the complex visual scenes in road traffic. Open access datasets for autonomous driving are primarily recorded with street vehicles and therefore only contain scenes in structured traffic areas but not in areas for cyclists and pedestrians. In this paper, we present a dataset for object detection and semantic segmentation that is specifically tailored to the particular challenges of autonomous robots in urban environments. Our dataset, recorded from the perspective of a bicycle, is based on real-world scenarios and different environments and includes a wide range of objects commonly encountered in urban environments. To ensure robustness and generality, our dataset includes hand-labeled annotations at a high level of detail, providing precise bounding boxes around each object of interest, and semantic segmentation masks at pixel level and comprehensive ground truth labels for training and evaluation. Our dataset will serve as a valuable resource for researchers, engineers and developers working on autonomous robotics, enabling advancements in object detection algorithms and fostering the development of safer and more efficient autonomous robot systems in urban scenarios.</p>