LiDAR and Multi-Sensor 3D Perception
Detecting and segmenting objects in 3D from LiDAR point clouds, cameras or both, mainly for driving and robotics.
Tier Grades
Tier A: Frontier AI labs
feasible
Driving companies and big labs lead 3D perception with large sensor fleets.
Tier B: Indonesian AI labs
stretch
An Indonesian lab can fine-tune on public LiDAR data but lacks fleet data.
Tier C: Institution units such as ITB
feasible
A 3-author team produced CenterPoint; a university unit can work on it.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
stretch
Client use, such as mining haul trucks or plantation machinery, needs sensor data collection first.
Where this comes from
Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.
Anchors
- ComputeTraining compute is not in the sources read.
- DataNot in the sources read; CenterPoint detects and tracks 3D objects as points in a point cloud.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 3 authors on the CenterPoint paper (arXiv 2006.11275).
- Time horizonYears.
- Shelf-life signalsEnd-to-end driving models absorb separate 3D detectors, Camera-only systems match LiDAR accuracy
- Pipeline roleTier A and driving firms train on fleet data; Tier C adapts models to local sensors; Tier D pilots 3D perception where clients have LiDAR.
Where this comes from
Team size is a proxy (author counts of exemplar papers), not a headcount.
Who and What
Exemplars
Observed actors
Linked models
BEVFusionvisionexemplar
CenterPointvisionexemplar
PointPillarsvisionexemplar
PointNetvisionexemplar
VoxelNeXtvisionproduced byTR3Dvisionproduced by
DSVTvisionproduced by
CenterFormervisionproduced byFocals Convvisionproduced by
TransFusionvisionproduced by
FCAF3Dvisionproduced byDETR3Dvisionproduced by
DD3Dvisionproduced byPGD (Probabilistic and Geometric Depth)visionproduced byImVoxelNetvisionproduced byFCOS3Dvisionproduced by
MonoFlexvisionproduced by
Group-Free-3Dvisionproduced by
PAConvvisionproduced by
CaDDNvisionproduced byPV-RCNN++visionproduced byVoxel R-CNNvisionproduced by
Objectron 3D object detectionvisionproduced byCylinder3Dvisionproduced by+17
SPVCNNvisionproduced byH3DNetvisionproduced bySSN (Shape Signature Networks)visionproduced by3DSSDvisionproduced bySMOKEvisionproduced by
ImVoteNetvisionproduced byPV-RCNNvisionproduced byPart-A2visionproduced byVoteNetvisionproduced byMinkUNetvisionproduced by
MVX-Netvisionproduced byPointRCNNvisionproduced byDGCNNvisionproduced byPointNet++visionproduced byPointNetvisionproduced byPV-RCNNvisionproduced byVoteNetvisionproduced byLinks
R&D TopicLiDAR and Multi-Sensor 3D Perception
ModelsPointNetStanford University, visionPointPillarsnuTonomy, visionCenterPointUniversity of Texas at Austin, visionBEVFusionMassachusetts Institute of Technology, vision+37 more
PointNetStanford University, visionPointPillarsnuTonomy, visionCenterPointUniversity of Texas at Austin, visionBEVFusionMassachusetts Institute of Technology, vision+37 moreUse Casesnone recorded
Who Runs Itnone recorded
Models 41
PV-RCNNThe Chinese University of Hong Kong, SenseTime, visionproduced byDETR3DMassachusetts Institute of Technology, Toyota Research Institute (TRI), visionproduced byImVoteNetMeta, Stanford University, visionproduced byPointNetStanford University, visionproduced byPV-RCNNThe Chinese University of Hong Kong, SenseTime, visionproduced byPV-RCNN++The Chinese University of Hong Kong, Max Planck Institute for Informatics, visionproduced bySPVCNNTsinghua University, Massachusetts Institute of Technology, visionproduced byVoteNetMeta, visionproduced byBEVFusionMassachusetts Institute of Technology, visionexemplarCenterFormerUniversity of Central Florida, TuSimple, visionproduced byCylinder3DThe Chinese University of Hong Kong, visionproduced byDD3DToyota Research Institute (TRI), visionproduced by+29
Focals ConvThe Chinese University of Hong Kong, Megvii, visionproduced byGroup-Free-3DUniversity of Science and Technology of China (USTC), Microsoft Research, visionproduced byImVoxelNetSamsung, visionproduced byMinkUNetStanford University, visionproduced byMVX-NetJohns Hopkins University, Apple, visionproduced byObjectron 3D object detectionGoogle, visionproduced byPointNetStanford University, visionexemplarPointNet++Stanford University, visionproduced byVoteNetMeta, visionproduced byVoxelNeXtThe Chinese University of Hong Kong, Megvii, visionproduced by3DSSDThe Chinese University of Hong Kong, visionproduced byCaDDNUniversity of Toronto, visionproduced byCenterPointUniversity of Texas at Austin, visionexemplarDGCNNMassachusetts Institute of Technology, visionproduced byDSVTPeking University, visionproduced byFCAF3DSamsung, visionproduced byFCOS3DThe Chinese University of Hong Kong, visionproduced byH3DNetUniversity of Texas at Austin, visionproduced byMonoFlexTsinghua University, visionproduced byPAConvThe University of Hong Kong, visionproduced byPart-A2The Chinese University of Hong Kong, visionproduced byPGD (Probabilistic and Geometric Depth)The Chinese University of Hong Kong, visionproduced byPointPillarsnuTonomy, visionexemplarPointRCNNThe Chinese University of Hong Kong, visionproduced bySMOKEZongmu Technology, visionproduced bySSN (Shape Signature Networks)The Chinese University of Hong Kong, visionproduced byTransFusionHong Kong University of Science and Technology, visionproduced byTR3Dvisionproduced byVoxel R-CNNvisionproduced byPapers
- PointNet: Deep Learning on Point Sets for 3D Classification and SegmentationCharles R. Qi et al., CVPR 2017 (per arXiv comment), 2016 founding
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object DetectionYin Zhou et al., arXiv preprint, 2017 founding
- BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View RepresentationZhijian Liu et al., ICRA 2023 (per arXiv comment), 2022 state of the art
- Center-based 3D Object Detection and TrackingTianwei Yin et al., Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognitio, 2020 state of the art
- PointPillars: Fast Encoders for Object Detection from Point CloudsAlex H. Lang et al., CVPR 2019 (per arXiv comment), 2018 state of the art
Where this comes from
The papers to read first: where the direction began, its current best, and a survey when one exists.
Sources
- arxiv.org/abs/2006.11275 https://arxiv.org/abs/2006.11275
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "LiDAR and Multi-Sensor 3D Perception", https://aurion.semai.id/topics/lidar-and-multi-sensor-3d-perception/, records of October 2026.