Pillar-Based 3D Object Detection from Point Cloud with Multiattention Mechanism

Figure 3

Visualization of our results. Qualitative results of the proposed method on the KITTI validation split. For each sample, the upper part is the image labeled with the predicted 3D bounding box, and the lower part is a representative view of the corresponding point clouds (best viewed when zoomed-in).