3D Dense Captioning on Nr3D
68.1C Score (0.5 IoU)APEIRIA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| APEIRIACategory=3D MLLMs2026.05 | 68.1 | — | — | — | — | |
| PVCap-Swin3DLoss=SCST, Backbone=Swin3D2026.07 | 61.61 | 33.46 | 27.49 | 60.55 | — | |
| LEGOCategory=3D MLLMs2026.05 | 61.4 | — | — | — | — | |
| PerLASegmenter=✓, #Params=119.76M, FLOP=326.762026.02 | 55.06 | 31.24 | 28.52 | 59.13 | — | |
| Fase3DSegmenter=✓, #Params=10.54M, FLOP=2.042026.02 | 54.91 | 30.24 | 26.48 | 57.14 | — | |
| PVCap-Swin3DLoss=MLE, Backbone=Swin3D2026.07 | 53.31 | 30.14 | 27.61 | 59.98 | — | |
| Fase3DSegmenter=✗, #Params=10.54M, FLOP=2.042026.02 | 52.89 | 29.31 | 26.14 | 56.41 | — | |
| LL3DACategory=3D MLLMs2026.05 | 51.2 | — | — | — | — | |
| LL3DASegmenter=✓, #Params=118.87M, FLOP=80.432026.02 | 51.18 | 28.75 | 25.91 | 56.61 | — | |
| Vote2Cap-DETR++Loss=SCST2026.07 | 47.62 | 28.41 | 25.63 | 54.77 | — | |
| V2C-DETR++Category=Specialist Methods2026.05 | 47.1 | — | — | — | — | |
| Vote2Cap-DETR++Loss=MLE2026.07 | 47.08 | 27.7 | 25.44 | 55.22 | — | |
| Vote2Cap-DETRLoss=SCST2026.07 | 45.53 | 26.88 | 25.43 | 54.76 | — | |
| Vote2Cap-DETRLoss=MLE2026.07 | 43.84 | 26.68 | 25.41 | 54.43 | — | |
| V2C-DETRCategory=Specialist Methods2026.05 | 43.8 | — | — | — | — | |
| D3NetLoss=SCST2026.07 | 38.42 | 22.22 | 24.74 | 54.37 | — | |
| 3DJCGLoss=MLE2026.07 | 38.06 | 22.82 | 23.77 | 52.99 | — | |
| Global Context Modeling (GCM) + Local Context Modeling (LCM)Data=2D+3D, Detector=VoteNet, CIDEr reward training=true2022.10 | 37.37 | 20.96 | 22.89 | 51.11 | 39.94 | |
| ContextualLoss=SCST2026.07 | 37.37 | 20.96 | 22.89 | 51.11 | — | |
| Global Context Modeling (GCM) + Local Context Modeling (LCM)Data=3D, Detector=VoteNet, CIDEr reward training=true2022.10 | 35.86 | 20.73 | 22.86 | 51.23 | 38.35 | |
| Global Context Modeling (GCM) + Local Context Modeling (LCM)Data=2D+3D, Detector=VoteNet, CIDEr reward training=false2022.10 | 35.26 | 20.42 | 22.77 | 50.78 | 39.29 | |
| ContextualLoss=MLE2026.07 | 35.26 | 20.42 | 22.77 | 50.78 | — | |
| REMANLoss=MLE2026.07 | 34.81 | 20.37 | 23.01 | 50.99 | — | |
| Global Context Modeling (GCM) + Local Context Modeling (LCM)Data=3D, Detector=VoteNet, CIDEr reward training=false2022.10 | 34.67 | 20.22 | 22.54 | 50.88 | 38.12 | |
| D3NetLoss=MLE2026.07 | 33.85 | 20.7 | 23.13 | 53.38 | — | |
| SpaCap3DData=2D+3D, Detector=VoteNet2022.10 | 33.71 | 19.92 | 22.61 | 50.5 | 38.11 | |
| SpaCap3dLoss=MLE2026.07 | 33.71 | 19.92 | 22.61 | 50.5 | — | |
| X-Trans2CapData=2D+3D, Detector=VoteNet2022.10 | 33.62 | 19.29 | 22.27 | 50 | 34.38 | |
| χ-Tran2CapLoss=SCST2026.07 | 33.62 | 19.29 | 22.27 | 50 | — | |
| SpaCap3DData=3D, Detector=VoteNet2022.10 | 31.43 | 18.98 | 22.24 | 49.79 | 33.17 | |
| X-Trans2CapData=3D, Detector=VoteNet2022.10 | 30.96 | 18.7 | 22.15 | 49.92 | 34.13 | |
| Scan2CapCategory=Specialist Methods2026.05 | 27.5 | — | — | — | — | |
| Scan2CapLoss=MLE2026.07 | 27.47 | 17.24 | 21.8 | 49.06 | — | |
| Scan2CapData=2D+3D, Detector=VoteNet2022.10 | 24.1 | 15.01 | 21.01 | 47.95 | 32.21 |