PAI3D: Painting Adaptive Instance-Prior for 3D Object Detection

3D object detection is a critical task in autonomous driving. Recently multi-modal fusion-based methods, which combine the complementary advantages of LiDAR and camera, have shown great performance improvements over mono-modal methods. However, so far, no methods attempted to utilize instance-level contextual image semantics guide detection. In this paper, we propose simple effective Painting Adaptive Instance-prior for (PAI3D) fuse flexibly with point cloud features. PAI3D sequential fusion framework. It first extracts semantic information from images, extracted information, including objects categorical label, point-to-object membership position, are then used augment each subsequent network improve performance. outperforms state-of-the-art large margin on nuScenes dataset, achieving 71.4 mAP 74.2 NDS test split. Our comprehensive experiments show that contribute most gain, works well any good-quality instance segmentation models modern encoders, making it strong candidate deployment vehicles.