3D Object Detection on SUN RGB-D (val)
69.7mAP@0.25Point-GCC
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Point-GCCTraining Protocol=Fine-tuning with pre-training, Backbone=TR3D+FF2023.05 | 69.7 | 54 | — | — | — | — | — | — | — | — | — | — | |
| TR3D+FFTraining Protocol=Self-supervised Pre-training (Baseline)2023.05 | 69.4 | 53.4 | — | — | — | — | — | — | — | — | — | — | |
| V-DETRProtocol=Single Run, TTA=true2023.08 | 68 | 51.1 | — | — | — | — | — | — | — | — | — | — | |
| Point-GCCTraining Protocol=Fine-tuning with pre-training, Backbone=TR3D2023.05 | 67.7 | 51 | — | — | — | — | — | — | — | — | — | — | |
| V-DETRProtocol=Single Run, TTA=false2023.08 | 67.5 | 50.4 | — | — | — | — | — | — | — | — | — | — | |
| V-DETRProtocol=Average 25x trials, TTA=true2023.08 | 67.5 | 50 | — | — | — | — | — | — | — | — | — | — | |
| TR3DTraining Protocol=Supervised Only2023.05 | 67.1 | 50.4 | — | — | — | — | — | — | — | — | — | — | |
| TR3DTraining Protocol=Self-supervised Pre-training (Baseline)2023.05 | 67.1 | 50.4 | — | — | — | — | — | — | — | — | — | — | |
| TR3DProtocol=Single Run2023.08 | 67.1 | 50.4 | — | — | — | — | — | — | — | — | — | — | |
| CAGroup3Dmulti-sensor=false2022.10 | 66.8 | 50.2 | — | — | — | — | — | — | — | — | — | — | |
| CAGroup3Dstat=max2023.05 | 66.8 | 50.2 | — | — | — | — | — | — | — | — | — | — | |
| CAGroup3DProtocol=Single Run2023.08 | 66.8 | 50.2 | — | — | — | — | — | — | — | — | — | — | |
| V-DETRProtocol=Average 25x trials, TTA=false2023.08 | 66.8 | 49.7 | — | — | — | — | — | — | — | — | — | — | |
| CAGroup3Dstat=average2023.05 | 66.4 | 49.5 | — | — | — | — | — | — | — | — | — | — | |
| CAGroup3DProtocol=Average 25x trials2023.08 | 66.4 | 49.5 | — | — | — | — | — | — | — | — | — | — | |
| TR3DProtocol=Average 25x trials2023.08 | 66.3 | 49.6 | — | — | — | — | — | — | — | — | — | — | |
| TR3DPresent at=ICIP, Paradigm=Fully Supervised2025.03 | 66.3 | 49.6 | — | — | — | — | — | — | — | — | — | — | |
| OctFormerbackbone=OctFormer, stat=max2023.05 | 66.2 | 50.6 | — | — | — | — | — | — | — | — | — | — | |
| OctFormerbackbone=OctFormer, stat=average2023.05 | 65.7 | 50.2 | — | — | — | — | — | — | — | — | — | — | |
| BrTRGB=true2022.10 | 65.4 | 48.1 | 82.8 | 88 | 40.5 | 79.7 | 33.4 | 40.6 | 67.4 | 71.1 | 55.7 | 93.5 | |
| SPGroup3DAggregation=Best2023.12 | 65.4 | 47.1 | — | — | — | — | — | — | — | — | — | — | |
| MTC-RCNNBackbone=PointNet++, Input=RGB + Point Cloud2022.04 | 65.3 | 48.6 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree (12, 512) with RepSurf-U²Backbone=RepSurf-U², #Params=23.61M, Infer Speed=195ms2022.05 | 64.9 | 47.7 | — | — | — | — | — | — | — | — | — | — | |
| TokenFusionPresented at=CVPR 22, multi-sensor=true2022.10 | 64.9 | 48.3 | — | — | — | — | — | — | — | — | — | — | |
| Point-GCCTraining Protocol=Fine-tuning with pre-training, Backbone=VoteNet+FF2023.05 | 64.9 | 41.3 | — | — | — | — | — | — | — | — | — | — | |
| SPGroup3DAggregation=Average (25 trials)2023.12 | 64.8 | 46.4 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree3D + PiMAEPre-trained=SUN RGB-D2023.03 | 64.6 | 46.2 | — | — | — | — | — | — | — | — | — | — | |
| VoteNet+FFTraining Protocol=Self-supervised Pre-training (Baseline)2023.05 | 64.5 | 39.2 | — | — | — | — | — | — | — | — | — | — | |
| ORACLEDescription=Upper bound (full annotation), Base detector=CAGroup3D2025.03 | 64.31 | 47.43 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree (6, 256) with RepSurf-UBackbone=RepSurf-U, #Params=11.50M, Infer Speed=150ms2022.05 | 64.3 | 45.9 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree (12, 512) with RepSurf-T²Backbone=RepSurf-T², #Params=23.60M, Infer Speed=194ms2022.05 | 64.2 | 47.1 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3DPresented at=ECCV'22, multi-sensor=false2022.10 | 64.2 | 48.9 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3Dstat=max2023.05 | 64.2 | 48.9 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3DTraining Protocol=Supervised Only2023.05 | 64.2 | 48.9 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3DProtocol=Single Run2023.08 | 64.2 | 48.9 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3DPresented at=ECCV'22, Aggregation=Best2023.12 | 64.2 | 48.9 | — | — | — | — | — | — | — | — | — | — | |
| RBGNetBackbone=PointNet++, Rays=66, Object Candidates=256, Input=Point Cloud2022.04 | 64.1 | 47.2 | — | — | — | — | — | — | — | — | — | — | |
| RBGNetPresented at=CVPR 22, multi-sensor=false2022.10 | 64.1 | 47.2 | — | — | — | — | — | — | — | — | — | — | |
| RBGNetProtocol=Single Run2023.08 | 64.1 | 47.2 | — | — | — | — | — | — | — | — | — | — | |
| RBGNetPresented at=CVPR 22, Aggregation=Best2023.12 | 64.1 | 47.2 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree (6, 256) with RepSurf-TBackbone=RepSurf-T, #Params=11.50M, Infer Speed=149ms2022.05 | 63.9 | 45.6 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3Dstat=average2023.05 | 63.8 | 48.2 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3DProtocol=Average 25x trials2023.08 | 63.8 | 48.2 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3DPresented at=ECCV'22, Aggregation=Average (25 trials)2023.12 | 63.8 | 48.2 | — | — | — | — | — | — | — | — | — | — | |
| FCAF3DPresent at=ECCV, Paradigm=Fully Supervised2025.03 | 63.8 | 48.2 | — | — | — | — | — | — | — | — | — | — | |
| RBGNetProtocol=Average 25x trials2023.08 | 63.6 | 46.3 | — | — | — | — | — | — | — | — | — | — | |
| RBGNetPresented at=CVPR 22, Aggregation=Average (25 trials)2023.12 | 63.6 | 46.3 | — | — | — | — | — | — | — | — | — | — | |
| HyperDet3DPresented at=CVPR 22, multi-sensor=false2022.10 | 63.5 | 47.3 | — | — | — | — | — | — | — | — | — | — | |
| HyperDet3DProtocol=Single Run2023.08 | 63.5 | 47.3 | — | — | — | — | — | — | — | — | — | — | |
| HyperDet3DPresented at=CVPR 222023.12 | 63.5 | 47.3 | — | — | — | — | — | — | — | — | — | — | |
| ImVoteNetBackbone=PointNet++, Input=RGB + Point Cloud2022.04 | 63.4 | — | — | — | — | — | — | — | — | — | — | — | |
| ImVoteNetBackbone=PointNet++, Input=RGB+Depth2022.05 | 63.4 | — | — | — | — | — | — | — | — | — | — | — | |
| Group-Free + YOLOSRGB=true2022.10 | 63.4 | 45.7 | 80.6 | 87.5 | 35 | 78.5 | 32.2 | 37.3 | 66.7 | 69.3 | 54.4 | 92.1 | |
| imVoteNetRGB=true2022.10 | 63.4 | — | 75.9 | 87.6 | 41.3 | 76.7 | 28.7 | 41.4 | 69.9 | 70.7 | 51.1 | 90.5 | |
| ImVoteNetPresented at=CVPR'20, multi-sensor=true2022.10 | 63.4 | — | — | — | — | — | — | — | — | — | — | — | |
| Group-freeBackbone=PointNet++, Input=Point Cloud2022.04 | 63 | 45.2 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree (6, 256)Backbone=PointNet++, #Params=11.49M, Infer Speed=149ms2022.05 | 63 | 45.2 | — | — | — | — | — | — | — | — | — | — | |
| Group-FreeRGB=false2022.10 | 63 | 45.2 | 80 | 87.8 | 32.5 | 79.4 | 32.6 | 36 | 66.7 | 70 | 53.8 | 91.1 | |
| Group-freePresented at=ICCV'21, multi-sensor=false2022.10 | 63 | 45.2 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree3DPre-trained=None2023.03 | 63 | 45.2 | — | — | — | — | — | — | — | — | — | — | |
| GroupFreestat=max2023.05 | 63 | 45.2 | — | — | — | — | — | — | — | — | — | — | |
| Group-FreeProtocol=Single Run2023.08 | 63 | 45.2 | — | — | — | — | — | — | — | — | — | — | |
| GroupFreePresented at=ICCV'21, Aggregation=Best2023.12 | 63 | 45.2 | — | — | — | — | — | — | — | — | — | — | |
| Group-Free + Faster R-CNNRGB=true2022.10 | 62.7 | 44.2 | 78 | 87.4 | 34.3 | 77.2 | 32.8 | 36.5 | 67 | 68.8 | 53.2 | 91.6 | |
| GroupFreestat=average2023.05 | 62.6 | 44.4 | — | — | — | — | — | — | — | — | — | — | |
| Group-FreeProtocol=Average 25x trials2023.08 | 62.6 | 44.4 | — | — | — | — | — | — | — | — | — | — | |
| GroupFreePresented at=ICCV'21, Aggregation=Average (25 trials)2023.12 | 62.6 | 44.4 | — | — | — | — | — | — | — | — | — | — | |
| VENetBackbone=PointNet++, Input=Point Cloud2022.04 | 62.5 | 39.2 | — | — | — | — | — | — | — | — | — | — | |
| VENetPresented at=ICCV'21, multi-sensor=false2022.10 | 62.5 | 39.2 | — | — | — | — | — | — | — | — | — | — | |
| VENet2023.05 | 62.5 | 39.2 | — | — | — | — | — | — | — | — | — | — | |
| VENetProtocol=Single Run2023.08 | 62.5 | 39.2 | — | — | — | — | — | — | — | — | — | — | |
| VENetPresented at=ICCV'212023.12 | 62.5 | 39.2 | — | — | — | — | — | — | — | — | — | — | |
| Group-Free + 3-dim RGBRGB=true2022.10 | 61.7 | 42 | 77.1 | 87.2 | 31.2 | 76.5 | 30.8 | 36.2 | 66.3 | 68.1 | 53 | 90.7 | |
| HGNetBackbone=GU-net, Input=Point Cloud2022.04 | 61.6 | — | — | — | — | — | — | — | — | — | — | — | |
| HGNetRGB=false2022.10 | 61.6 | — | 78 | 84.5 | 35.7 | 75.2 | 34.3 | 37.6 | 61.7 | 65.7 | 51.6 | 91.1 | |
| HGNetPresented at=CVPR 20, multi-sensor=false2022.10 | 61.6 | — | — | — | — | — | — | — | — | — | — | — | |
| HGNet2023.05 | 61.6 | — | — | — | — | — | — | — | — | — | — | — | |
| HGNetProtocol=Single Run2023.08 | 61.6 | — | — | — | — | — | — | — | — | — | — | — | |
| HGNetPresented at=CVPR 202023.12 | 61.6 | — | — | — | — | — | — | — | — | — | — | — | |
| Point-GCCTraining Protocol=Fine-tuning with pre-training, Backbone=VoteNet, Implementation=Stronger MMDetection3D2023.05 | 61.3 | 37.7 | — | — | — | — | — | — | — | — | — | — | |
| GroupFree3D + PiMAE (from scratch)Pre-trained=None2023.03 | 61.2 | 44.7 | — | — | — | — | — | — | — | — | — | — | |
| VoteNet + HPE_MLPImplementation framework=MMDetection3D2026.03 | 61.2 | — | 77.1 | 85 | 34.3 | 78.5 | 28.5 | 35 | 64.1 | 67.7 | 50.5 | 91.1 | |
| BRNetBackbone=PointNet++, Input=Point Cloud2022.04 | 61.1 | 43.7 | — | — | — | — | — | — | — | — | — | — | |
| BRNetBackbone=PointNet++2022.05 | 61.1 | 43.7 | — | — | — | — | — | — | — | — | — | — | |
| BRNetPresented at=CVPR'21, multi-sensor=false2022.10 | 61.1 | 43.7 | — | — | — | — | — | — | — | — | — | — | |
| BRNet2023.05 | 61.1 | 43.7 | — | — | — | — | — | — | — | — | — | — | |
| BRNetProtocol=Single Run2023.08 | 61.1 | 43.7 | — | — | — | — | — | — | — | — | — | — | |
| BRNetPresented at=CVPR'212023.12 | 61.1 | 43.7 | — | — | — | — | — | — | — | — | — | — | |
| BRNet2026.03 | 61.1 | — | 76.2 | 86.9 | 29.7 | 77.4 | 29.6 | 35.9 | 65.9 | 66.4 | 51.8 | 91.3 | |
| PonderTraining Protocol=Fine-tuning with pre-training, Backbone=VoteNet2023.05 | 61 | 36.6 | — | — | — | — | — | — | — | — | — | — | |
| Ponder-RGBDDetection Model=VoteNet, Pre-training Type=Rendering, Pre-training Epochs=1002023.10 | 61 | 36.6 | — | — | — | — | — | — | — | — | — | — | |
| DepthContrastTraining Protocol=Fine-tuning with pre-training, Backbone=VoteNet2023.05 | 60.4 | 35.4 | — | — | — | — | — | — | — | — | — | — | |
| DepthContrastDetection Model=VoteNet, Pre-training Type=Contrast, Pre-training Epochs=10002023.10 | 60.4 | 35.4 | — | — | — | — | — | — | — | — | — | — | |
| IAEDetection Model=VoteNet, Pre-training Type=Masked Auto-Encoding, Pre-training Epochs=10002023.10 | 60.4 | 36 | — | — | — | — | — | — | — | — | — | — | |
| CPDet3DPresent at=-, Paradigm=Sparse Supervised2025.03 | 60.2 | 43.3 | — | — | — | — | — | — | — | — | — | — | |
| H3DNetBackbone=4xPointNet++, Input=Point Cloud2022.04 | 60.1 | 39 | — | — | — | — | — | — | — | — | — | — | |
| H3DNetBackbone=4×PointNet++, Infer Speed=266ms2022.05 | 60.1 | 39 | — | — | — | — | — | — | — | — | — | — | |
| H3DNetRGB=false2022.10 | 60.1 | 39 | 73.8 | 85.6 | 31 | 76.7 | 29.6 | 33.4 | 65.5 | 66.5 | 50.8 | 88.2 | |
| H3DNetPresented at=ECCV'20, multi-sensor=false2022.10 | 60.1 | 39 | — | — | — | — | — | — | — | — | — | — | |
| H3DNet2023.05 | 60.1 | 39 | — | — | — | — | — | — | — | — | — | — |