The trend in autonomous driving is moving towards end-to-end models that are solving tasks jointly through fusion of sensing data from different modalities at the temporal level rather than using a frame-by-frame detection method.