0d18ab3b5fabfa6fe47c62e711af02f0-Supplemental-Conference.pdf

Inpractice,astronger 3D object detection method as ourDeepInteraction model is expected to reduce the potential accidents of self-driving cars. This improves the safety and reliability of autonomous driving. A.3 Limitations All the components for multi-modal fusion in ourDeepInteraction have no preference to any per-modalrepresentations. We will explore how to generate initial queries from both modalities (i.e., LiDAR'sbird-eyes-viewandcamera'sfront-view). The bounding boxesofground-truth and predictions are inthe color blue and green respectively.