3D Question Answering on ScanQA w/o objects (test)
30.87EM@1BridgeQA
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| BridgeQAUse question to match 2D views at test time=true2024.02 | 30.87 | — | — | — | — | — | — | — | — | — | |
| BridgeQA2024.02 | 30.82 | — | — | — | — | — | — | — | — | — | |
| BridgeQA2024.02 | 30.82 | — | 34.41 | — | — | 17.74 | 41.18 | 15.6 | 79.34 | — | |
| DSPNetPre-trained=true, LLMs-based=false, Extra dataset=true2025.03 | 23.8 | — | — | — | — | — | — | — | — | — | |
| DSPNetPre-trained=false2025.03 | 23.8 | 56.1 | — | — | — | 15.7 | 35.1 | 14.3 | 69.6 | — | |
| GPSPre-trained=true, LLMs-based=false, Extra dataset=true2025.03 | 23.5 | — | — | — | — | — | — | — | — | — | |
| 3D-VisTAObject Proposals=Mask3D, Pre-training status=Pre-trained2023.08 | 23 | 53.5 | — | — | — | 11.9 | 32.8 | 12.9 | 62.6 | — | |
| 3D-VisTA2024.02 | 23 | — | — | — | — | — | — | — | — | — | |
| 3D-VisTA2024.02 | 23 | — | — | — | — | 11.9 | 32.8 | 12.9 | 62.6 | — | |
| 3D-VisTA2024.05 | 23 | — | 30.2 | — | — | — | — | 12.9 | 62.6 | — | |
| 3D-VisTAPre-trained=true2025.03 | 23 | 53.5 | — | — | — | 11.9 | 32.8 | 12.9 | 62.6 | — | |
| 3DGraphQAPre-trained=false2025.03 | 22.3 | 56.1 | — | — | — | 12.9 | 33 | 13.6 | 62.9 | — | |
| RandomImage+Oscar (real)Image Source=Random Image, Model=Oscar, Feature Type=real2021.12 | 21.58 | 49.85 | 24.86 | 16.16 | 13.29 | 0 | 28.99 | 10.99 | 54.62 | 8.62 | |
| 3DVLP (Jin et al. 2023)2024.02 | 21.56 | — | — | — | — | — | — | — | — | — | |
| 3DVLP2024.02 | 21.56 | — | 31.48 | — | — | 15.84 | 31.79 | 13.13 | 63.4 | — | |
| Multi-CLIPPre-trained=true2025.03 | 21.5 | — | — | — | — | 12.9 | 32.6 | 13.4 | 63.2 | — | |
| Multi-CLIP2024.02 | 21.48 | — | — | — | — | — | — | — | — | — | |
| Multi-CLIP2024.02 | 21.48 | — | 32.69 | — | — | 12.87 | 32.61 | 13.36 | 63.2 | — | |
| CLIP-Guided2024.02 | 21.37 | — | — | — | — | — | — | — | — | — | |
| CLIP-Guided2024.02 | 21.37 | — | 32.7 | — | — | 11.73 | 32.41 | 13.28 | 62.83 | — | |
| ScanQA (multiple)Variant=multiple2021.12 | 21.3 | 53.05 | 31.14 | 21.2 | 15.81 | 11.18 | 31.62 | 12.82 | 60.95 | 11.68 | |
| ScanQA2021.12 | 20.9 | 54.11 | 30.68 | 21.2 | 15.81 | 10.75 | 31.09 | 12.59 | 60.24 | 11.29 | |
| ScanQA (single)Variant=single2021.12 | 20.9 | 54.11 | 30.68 | 21.2 | 15.81 | 10.75 | 31.09 | 12.59 | 60.24 | 11.29 | |
| ScanQAObject Proposals=Mask3D2023.08 | 20.9 | 54.1 | — | — | — | 10.8 | 31.1 | 12.6 | 60.2 | — | |
| ScanQA2024.02 | 20.9 | — | — | — | — | — | — | — | — | — | |
| ScanQA2024.02 | 20.9 | — | 30.68 | — | — | 10.75 | 31.09 | 12.59 | 60.24 | — | |
| ScanQA2024.05 | 20.9 | — | 30.7 | — | — | — | — | 12.6 | 60.2 | — | |
| ScanQAPre-trained=false2025.03 | 20.9 | 54.1 | — | — | — | 10.8 | 31.1 | 12.6 | 60.2 | — | |
| RandomImage+MCANsampling strategy=RandomImage, 2D-QA model=MCAN2021.12 | 20.82 | 51.23 | 26.29 | 17.9 | 14.27 | 9.66 | 29.23 | 11.54 | 55.64 | 8.87 | |
| RandomImage+MCAN (real)Image Source=Random Image, Model=MCAN, Feature Type=real2021.12 | 20.82 | 51.23 | 26.29 | 17.9 | 14.27 | 9.66 | 29.23 | 11.54 | 55.64 | 8.87 | |
| Image+MCANObject Proposals=Mask3D2023.08 | 20.8 | 51.2 | — | — | — | 9.7 | 29.2 | 11.5 | 55.6 | — | |
| Image+MCANPre-trained=false2025.03 | 20.8 | 51.2 | — | — | — | 9.7 | 29.2 | 11.5 | 55.6 | — | |
| 3D-VisTA (scratch)Object Proposals=Mask3D, Pre-training status=From scratch2023.08 | 20.4 | 51.5 | — | — | — | 8.7 | 29.6 | 11.6 | 55.7 | — | |
| 3D-VisTAPre-trained=false2025.03 | 20.4 | 51.5 | — | — | — | 8.7 | 29.6 | 11.6 | 55.7 | — | |
| RandomImage+MCAN (mesh)Image Source=Random Image, Model=MCAN, Feature Type=mesh2021.12 | 20.34 | 51.2 | 24.72 | 16.93 | 14.04 | 7.55 | 28.1 | 10.95 | 53.41 | 8.61 | |
| ScanQA w/o multiview2024.02 | 20.05 | — | — | — | — | — | — | — | — | — | |
| PQ3DTraining Protocol=unified joint training2024.05 | 20 | — | 36.1 | — | — | — | — | 13.9 | 65.2 | — | |
| PQ3DPre-trained=false, LLMs-based=false, Extra dataset=true2025.03 | 20 | — | — | — | — | — | — | — | — | — | |
| RandomImage+Oscar (mesh)Image Source=Random Image, Model=Oscar, Feature Type=mesh2021.12 | 19.91 | 48.74 | 23.02 | 13.24 | 12.91 | 0.64 | 26.94 | 10.17 | 49.83 | 7.45 | |
| TopDownImage+OscarImage Source=Top-Down Image, Model=Oscar2021.12 | 19.13 | 48.06 | 21.93 | 11.66 | 14.64 | 0.7 | 25.98 | 9.67 | 47.37 | 6.93 | |
| ScanRefer+MCAN (e2e)architecture=end-to-end2021.12 | 19.04 | 49.7 | 26.98 | 16.17 | 11.28 | 7.82 | 28.61 | 11.38 | 53.41 | 10.63 | |
| ScanRefer+MCAN (e2e)Image Source=ScanRefer, Model=MCAN, Setup=end-to-end2021.12 | 19.04 | 49.7 | 26.98 | 16.17 | 11.28 | 7.82 | 28.61 | 11.38 | 53.41 | 10.63 | |
| ScanRefer+MCANObject Proposals=Mask3D2023.08 | 19 | 49.7 | — | — | — | 7.8 | 28.6 | 11.4 | 53.4 | — | |
| ScanRefer+MCANPre-trained=false2025.03 | 19 | 49.7 | — | — | — | 7.8 | 28.6 | 11.4 | 53.4 | — | |
| VoteNet+MCANdetector=VoteNet, 2D-QA model=MCAN2021.12 | 18.15 | 48.56 | 29.63 | 17.8 | 11.57 | 7.1 | 29.12 | 11.68 | 53.34 | 10.36 | |
| VoteNet+MCANImage Source=VoteNet, Model=MCAN2021.12 | 18.15 | 48.56 | 29.63 | 17.8 | 11.57 | 7.1 | 29.12 | 11.68 | 53.34 | 10.36 | |
| ScanRefer+MCAN (pipeline)architecture=pipeline2021.12 | 16.47 | 49.05 | 18.71 | 10.98 | 16.53 | 0.76 | 22.45 | 8.76 | 40.81 | 6.41 | |
| ScanRefer+MCAN (pipeline, real)Image Source=ScanRefer, Model=MCAN, Setup=pipeline, Feature Type=real2021.12 | 16.47 | 49.05 | 18.71 | 10.97 | 16.53 | 0.76 | 22.45 | 8.76 | 40.81 | 6.41 | |
| ScanRefer+MCAN (pipeline, mesh)Image Source=ScanRefer, Model=MCAN, Setup=pipeline, Feature Type=mesh2021.12 | 16.39 | 47.76 | 18.28 | 10.42 | 15.03 | 0.71 | 22.07 | 8.62 | 40.19 | 6.03 | |
| PQ3DTraining Protocol=single dataset2024.05 | 16.1 | — | 30.5 | — | — | — | — | 12.1 | 56 | — | |
| TopDownImage+MCANImage Source=Top-Down Image, Model=MCAN2021.12 | 14.39 | 45.94 | 15.7 | 8.86 | 12.56 | 0.62 | 19.26 | 7.71 | 34.59 | 5.22 |