Abstract: Recent progress on multi-modal 3D object detection has featured BEV (Bird-Eye-View) based fusion, which effectively unifies both LiDAR point clouds and camera images in a shared BEV space.
Abstract: Existing audio-visual event localization (AVE) handles manually trimmed videos with only a single instance in each of them. However, this setting is unrealistic as natural videos often ...