This is a paper and code list of some awesome 3D detection methods. We mainly collect LiDAR-involved methods in autonomous driving. It is worth noticing that we include both official and unofficial codes for each paper.
2021.1.11 p.m. Add SA-Det3D: Self-Attention Based Context-Aware 3D Object Detection .
2020.12.08 p.m. Add CIA-SSD: An IoU-Aware Single-Stage Object Detector .
2020.12.07 p.m. Add CVCNet which proposes a new multi-view fusion methods and cross-view consistency loss.
2020.11.25 p.m. Add **DA-PointRCNN
| Title | Pub. | Input |
|---|---|---|
| MV3D (Multi-View 3D Object Detection Network for Autonomous Driving) | CVPR2017 | I+L |
| F-PointNet (Frustum PointNets for 3D Object Detection from RGB-D Data) code | CVPR2018 | I+L |
| VoxelNet (VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection) | CVPR2018 | L |
| PIXOR (PIXOR: Real-time 3D Object Detection from Point Clouds) code | CVPR2018 | L |
| AVOD (Joint 3D Proposal Generation and Object Detection from View Aggregation) code | IROS2018 | I+L |
| ContFusion (Deep Continuous Fusion for Multi-Sensor 3D Object Detection) | ECCV2018 | I+L |
| SECOND (SECOND: Sparsely Embedded Convolutional Detection) code | Sensors 2018 | L |
| Complex-YOLO (Complex-YOLO: Real-time 3D Object Detection on Point Clouds) code | Axiv2018 | L |
| FBF（Fusing Bird’s Eye View LIDAR Point Cloud and Front View Camera Image for Deep Object Detection）code | Arxiv2018 | I+L |
| RoarNet (RoarNet: A Robust 3D Object Detection based on Region Approximation Refinement) code | IV2019 | I+L |
| PVCNN (Point-Voxel CNN for Efficient 3D Deep Learning) code | NIPS2019 | L |
| MMF(Multi-Task Multi-Sensor Fusion for 3D Object Detection) code | CVPR2019 | I+L |
| PointPillars (PointPillars: Fast Encoders for Object Detection from Point Clouds) code | CVPR2019 | L |
| Point RCNN (PointRCNN: 3D Object Proposal Generation and Detection from Point Cloud) code | CVPR2019 | L |
| LaserNet (LaserNet: An Efficient Probabilistic 3D Object Detector for Autonomous Driving) | CVPR2019 | L |
| LaserNet++ (Sensor Fusion for Joint 3D Object Detection and Semantic Segmentation) | CVPR2019 | I+L |
| **Fast PointRCNN **(Fast PointRCNN) | ICCV2019 | L |
| STD (STD: Sparse-to-Dense 3D Object Detector for Point Cloud) | ICCV2019 | L |
| VoteNet (Deep Hough Voting for 3D Object Detection in Point Clouds) code | ICCV2019 | L |
| MVX-Net (MVX-Net: Multimodal VoxelNet for 3D Object Detection) code | ICRA2019 | I+L |
| Patchs (Patch Refinement - Localized 3D Object Detection) | Arxiv2019 | L |
| StarNet (StarNet: Targeted Computation for Object Detection in Point Clouds) code | Arxiv2019 | L |
| F-ConvNet (Frustum ConvNet: Sliding Frustums to Aggregate Local Point-Wise Features for Amodal 3D Object Detection) | IROS2019 | I+L |
| PI-RCNN（An Efficient Multi-sensor 3D Object Detector with Point-based Attentive Cont-conv Fusion Module） | AAAI2020 | I+L |
| TANet (TANet: Robust 3D Object Detection from Point Clouds with Triple Attention) code | AAAI2020 | L |
| MVF (End-to-end multi-view fusion for 3d object detection in lidar point clouds) code | ICRL2020 | L |
| SegVoxelNet (SegVoxelNet: Exploring Semantic Context and Depth-aware Features for 3D Vehicle Detection from Point Cloud) | ICRA2020 | L |
| Voxel-FPN (Voxel-FPN: multi-scale voxel feature aggregation in 3D object detection from point clouds) | Sensors 2020 | L |
| AA3D (Adaptive and Azimuth-Aware Fusion Network of Multimodal Local Features for 3D Object Detection) | Neurocomputing2020 | I+L |
| Part A^2 (Part-A^ 2 Net: 3D Part-Aware and Aggregation Neural Network for Object Detection from Point Cloud) code | TPAMI2020 | L |
| PV-RCNN (PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection) code | CVPR2020 | L |
| 3D SSD (3DSSD: Point-based 3D Single Stage Object Detector) code | CVPR2020 | L |
| Associate-3Ddet (Associate-3Ddet: Perceptual-to-Conceptual Association for 3D Point Cloud Object Detection) code | CVPR2020 | L |
| HVNet (HVNet: Hybrid Voxel Network for LiDAR Based 3D Object Detection) code | CVPR2020 | L |
| ImVoteNet (ImVoteNet: Boosting 3D Object Detection in Point Clouds with Image Votes) | CVPR2020 | I+L |
| Point GNN (Point-GNN: Graph Neural Network for 3D Object Detection in a Point Cloud) | CVPR2020 | L |
| SA-SSD (Structure Aware Single-stage 3D Object Detection from Point Cloud) code | CVPR2020 | L |
| (What You See is What You Get: Exploiting Visibility for 3D Object Detection) | CVPR2020 | L |
| DOPS (DOPS: Learning to Detect 3D Objects and Predict their 3D Shapes) | CVPR2020 | L |
| 3D IoU-Net (3D IoU-Net: IoU Guided 3D Object Detector for Point Clouds) | Arxiv2020 | L |
| 3D CVF (3D-CVF: Generating Joint Camera and LiDAR Features Using Cross-View Spatial Feature Fusion for 3D Object Detection) | ECCV2020 | I+L |
| HotSpotNet (Object as Hotspots: An Anchor-Free 3D Object Detection Approach via Firing of Hotspots) | ECCV2020 | L |
| EPNet: (EPNet: Enhancing Point Features with Image Semantics for 3D Object Detection) code | ECCV2020 | I+L |
| WS3D (Weakly Supervised 3D Object Detection from Lidar Point Cloud) code | ECCV2020 | L |
| Pillar-OD Pillar-based Object Detection for Autonomous Driving code | ECCV2020 | L |
| SSN (SSN: Shape Signature Networks for Multi-class Object Detection from Point Clouds) | Arxiv2020 | L |
| CenterPoint (Center-based 3D Object Detection and Tracking) code | Arxiv2020 | L |
| AFDet (AFDet: Anchor Free One Stage 3D Object Detection) | Waymo2020 | L |
| LGR-Net (Local Grid Rendering Networks for 3D Object Detection in Point Clouds) | arxiv2020.07 | L |
| CenterNet3D (CenterNet3D:An Anchor free Object Detector for Autonomous Driving)code | arxiv2020.07 | L |
| RCD (Range Conditioned Dilated Convolutions for Scale Invariant 3D Object Detection) | arxiv2020.06 | L |
| VS3D (Weakly Supervised 3D Object Detection from Point Clouds) code | ACM MM2020 | I+L |
| LC-MV (Multi-View Fusion of Sensor Data for Improved Perception and Prediction in Autonomous Driving) | CoRL2020 | I+L |
| RangeRCNN (RangeRCNN: Towards Fast and Accurate 3D Object Detection with Range Image Representation) | arxiv2020.09 | L |
| MVAF-Net (Multi-View Adaptive Fusion Network for 3D Object Detection) | arxiv2020.11 | I+L |
| CADNet (Context-Aware Dynamic Feature Extraction for 3D Object Detection in Point Clouds) | arxiv2020.07 | L |
| DA-PointRCNN (A Density-Aware PointRCNN for 3D Objection Detection in Point Clouds) | axiv2020.09 | L |
| CVCNet(Every View Counts: Cross-View Consistency in 3D Object Detection with Hybrid-Cylindrical-Spherical Voxelization) | NIPS2020 | L |
| CIA-SSD（CIA-SSD: Confident IoU-Aware Single-Stage Object Detector From Point Cloud）code | AAAI2021 | L |
| IAAY（It's All Around You: Range-Guided Cylindrical Network for 3D Object Detection） | arxiv2020 | L |
| SA-Det3D (Self-Attention Based Context-Aware 3D Object Detection)code | arxiv2020 | L |
| To be continued... |
-
mmdetection3d in pytorch
Methods supported: SECOND, PointPillars, FreeAnchor, VoteNet, Part-A2, MVXNet
Benchmark supported: KITTI, nuScenes, Lyft, ScanNet, SUNRGBD
-
OpenPCDet: An open source project for LiDAR-based 3D scene perception in Pytorch.
Methods supported : PointPillars, SECOND, Part A^2, PV-RCNN, PointRCNN(ongoing).
Benchmark supported: KITTI, Waymo (ongoing).
-
Det3d: A general 3D Object Detection codebase in PyTorch.
Methods supported : PointPillars, SECOND, PIXOR.
Benchmark supported: KITTI, nuScenes, Lyft.
-
second.pytorch: SECOND detector in Pytorch.
Methods supported : PointPillars, SECOND.
Benchmark supported: KITTI, nuScenes.
-
CenterPoint: "Center-based 3D Object Detection and Tracking" in Pytorch.
Methods supported : CenterPoint-Pillar, Center-Voxel.
Benchmark supported: nuScenes，Waymo.
-
SA-SSD: "SA-SSD: Structure Aware Single-stage 3D Object Detection from Point Cloud" in pytorch
Methods supported : SA-SSD.
Benchmark supported: KITTI.
-
3DSSD: "Point-based 3D Single Stage Object Detector " in Tensorflow.
Methods supported : 3DSSD, PointRCNN, STD (ongoing).
Benchmark supported: KITTI, nuScenes (ongoing).
-
Point-GNN: "Point-GNN: Graph Neural Network for 3D Object Detection in a Point Cloud" in Tensorflow.
Methods supported : Point-GNN.
Benchmark supported: KITTI.
-
TANet: "TANet: Robust 3D Object Detection from Point Clouds with Triple Attention" in Pytorch.
Methods supported : TANet (PointPillars, Second).
Benchmark supported: KITTI.
-
Complex-YOLOv4-pytorch: " Complex-YOLO: Real-time 3D Object Detection on Point Clouds)" in pytorch.
Methods supported : YOLO
Benchmark supported: KITTI.
-
EPNet: "EPNet: Enhancing Point Features with Image Semantics for 3D Object Detection "
Methods supported: EPNet
Benchmark supported: KITTI, SUN-RGBD
-
Super Fast and Accurate 3D Detector:"Super Fast and Accurate 3D Object Detection based on 3D LiDAR Point Clouds"
Benchmark supported: KITTI
(reference: https://mp.weixin.qq.com/s/3mpbulAgiwi5J66MzPNpJA from WeChat official account: "CNNer")
-
KITTI
Website: http://www.cvlibs.net/datasets/kitti/raw_data.php
Paper: http://www.cvlibs.net/publications/Geiger2013IJRR.pdf
-
Waymo
Website: https://waymo.com/open
-
NuScenes
Website: https://www.nuscenes.org/
-
Lyft
Website: https://level5.lyft.com/
-
Audi autonomous driving dataset
Website: http://www.a2d2.audi
-
Apollo
Website: http://apolloscape.auto/