The cases in the first to third rows of the figure illustrate the impact of ego-vehicle turning on the prediction of the pedestrian crossing. The fourth column shows that when the model observes a pedestrian during the ego-vehicle turn, the pedestrian’s bounding box trajectory is similar to the samples moving rapidly to the right, making it easy to confuse the model with a fast-moving pedestrian trying to cross. Note that red bounding boxes indicate pedestrian samples that cross, while green ones indicate those that do not.