Homogeneous Multi-modal Feature Fusion and Interaction for 3D Object Detection

Multi-modal 3D object detection has been an active research topic in autonomous driving. Nevertheless, it is non-trivial to explore the cross-modal feature fusion between sparse points and dense 2D pixels. Recent approaches either fuse image features with point cloud that are projected onto plane or combine These often suffer from severe information loss, thus causing sub-optimal performance. To address these problems, we construct homogeneous structure images avoid projective loss by transforming camera into LiDAR space. In this paper, propose a multi-modal interaction method (HMFI) for detection. Specifically, first design voxel lifter module (IVLM) lift space generate features. Then, voxelized different regions introducing self-attention based query mechanism (QFM). Next, (VFIM) enforce consistency of semantic identical objects representations, which can provide object-level alignment guidance strengthen discriminative ability complex backgrounds. We conduct extensive experiments on KITTI Waymo Open Dataset, proposed HMFI achieves better performance compared state-of-the-art methods. Particularly, cyclist benchmark, surpasses all published algorithms large margin.