Unreliable classification of road objects and poor quality of training labels lead to unsafe automated driving decisions. Improving perception accuracy ensures robust vehicle operation in complex real-world scenarios.