Object as Hotspots: An Anchor-Free 3D Object Detection Approach via Firing of Hotspots

Abstract

Accurate 3D object detection in LiDAR based point clouds suffers from the challenges of data sparsity and irregularities. Existing methods strive to organize the points regularly, e.g. voxelize, pass them through a designed 2D/3D neural network, and then define object-level anchors that predict offsets of 3D bounding boxes using collective evidences from all the points on the objects of interest. Contrary to the state-of-the-art anchor-based methods, based on the very nature of data sparsity, we observe that even points on an individual object part are informative about semantic information of the object. We thus argue in this paper for an approach opposite to existing methods using object-level anchors. Inspired by compositional models, which represent an object as parts and their spatial relations, we propose to represent an object as composition of its interior non-empty voxels, termed hotspots, and the spatial relations of hotspots. This gives rise to the representation of Object as Hotspots (OHS). Based on OHS, we further propose an anchor-free detection head with a novel ground truth assignment strategy that deals with inter-object point-sparsity imbalance to prevent the network from biasing towards objects with more points. Experimental results show that our proposed method works remarkably well on objects with a small number of points. Notably, our approach ranked 1st on KITTI 3D Detection Benchmark for cyclist and pedestrian detection, and achieved state-of-the-art performance on NuScenes 3D Detection Benchmark.

arXiv commentsmall affiliation fix

References in corpus (16)

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Decoupled Weight Decay Regularization
• FCOS: Fully Convolutional One-Stage Object Detection
• ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
• Deep Continuous Fusion for Multi-Sensor 3D Object Detection
• Point-Voxel CNN for Efficient 3D Deep Learning
• Class-balanced Grouping and Sampling for Point Cloud 3D Object Detection
• End-to-End Learning of Geometry and Context for Deep Stereo Regression
• Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds
• Frustum ConvNet: Sliding Frustums to Aggregate Local Point-Wise Features for Amodal 3D Object Detection
• Center-based 3D Object Detection and Tracking
• STD: Sparse-to-Dense 3D Object Detector for Point Cloud
• Voxel-FPN: multi-scale voxel feature aggregation in 3D object detection from point clouds
• Learning a Hierarchical Compositional Shape Vocabulary for Multi-class Object Representation
• IoU Loss for 2D/3D Object Detection
• What You See is What You Get: Exploiting Visibility for 3D Object Detection

Cited by in corpus (5)

• AFDet: Anchor Free One Stage 3D Object Detection
• Deep Learning for 3D Point Clouds: A Survey
• PV-RCNN++: Point-Voxel Feature Set Abstraction With Local Vector Representation for 3D Object Detection
• SE-SSD: Self-Ensembling Single-Stage Object Detector From Point Cloud
• Learning to Reconstruct and Segment 3D Objects