3D Instance Segmentation on ScanNet200
63.7mAP@0.5ESAM
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| ESAMType=Online, 3D GT=Yes2026.01 | 63.7 | — | — | 42.2 | 79.6 | — | — | — | — | |
| 3AMType=Online, 3D GT=No2026.01 | 59.7 | — | — | 47.3 | 75.3 | — | — | — | — | |
| SAM2ObjectType=Offline, 3D GT=Yes2026.01 | 52.7 | — | — | 34 | 70.3 | — | — | — | — | |
| SegDINO3DInput Modality=P & I & D & C2026.03 | 52.4 | — | — | — | 58.6 | — | — | — | 40.2 | |
| SAI3DType=Offline, 3D GT=Yes2026.01 | 47.2 | — | — | 28.2 | 67.9 | — | — | — | — | |
| SegVGGTInput Modality=I2026.03 | 45.7 | — | — | — | 53.7 | — | — | — | 31.9 | |
| ODINInput Modality=I & D & C2026.03 | 45.3 | — | — | — | 53.1 | — | — | — | 31.5 | |
| Open3DISType=Offline, 3D GT=Yes2026.01 | 43.1 | — | — | 34.6 | 48.5 | — | — | — | — | |
| OV-3DIS3D mask proposal=2D+3D, evaluation protocol=Top-k2026.04 | 41.4 | — | — | — | 45.3 | 34.5 | 33.1 | 30.7 | 32.7 | |
| Relation3DInput Modality=P2026.03 | 41.2 | — | — | — | 45.6 | — | — | — | 31.6 | |
| OneFormer3DInput Modality=P2026.03 | 40.9 | — | — | — | 44.6 | — | — | — | 30.2 | |
| Mask3D3D mask proposal=3D, setting=closed-vocabulary2026.04 | 39.8 | — | — | — | 21.7 | 17.9 | — | — | 26.9 | |
| SGIFormer-LInput Modality=P2026.03 | 39.4 | — | — | — | 44.2 | — | — | — | 29.2 | |
| CAMFusion3D mask proposal=3D, evaluation protocol=Top-k2026.04 | 39.3 | — | — | — | 46.2 | 31.8 | 29.1 | 22.2 | 27.6 | |
| MAFTInput Modality=P2026.03 | 38.2 | — | — | — | 43.3 | — | — | — | 29.2 | |
| OV-3DIS3D mask proposal=3D, evaluation protocol=Top-k2026.04 | 37.6 | — | — | — | 42.8 | 33 | 25.3 | 28.1 | 29 | |
| QueryFormerInput Modality=P2026.03 | 37.1 | — | — | — | 43.4 | — | — | — | 28.1 | |
| Mask3DInput Modality=P2026.03 | 37 | — | — | — | 42.3 | — | — | — | 27.4 | |
| Mask3DSetting=Fully-sup2023.12 | 36.2 | — | — | 26.9 | 41.4 | 39.8 | 17.9 | 21.7 | — | |
| Mask3DMask supervision=supervised, Semantic supervision=supervised2024.11 | 36.2 | — | — | 26.9 | 41.4 | — | — | — | — | |
| SAM3DType=Online, 3D GT=Yes2026.01 | 35.7 | — | — | 20.2 | 55.5 | — | — | — | — | |
| CAMFusion3D mask proposal=3D, evaluation protocol=Top-12026.04 | 35 | — | — | — | 41 | 27.4 | 25.6 | 20.4 | 24.4 | |
| SAMPro3DType=Offline, 3D GT=Yes2026.01 | 32.8 | — | — | 18 | 56.1 | — | — | — | — | |
| ISBNetSetting=Fully-sup2023.12 | 32.7 | — | — | 24.5 | 37.6 | 38.6 | 12.5 | 20.5 | — | |
| OV-3DIS3D mask proposal=2D+3D, evaluation protocol=Top-12026.04 | 32.5 | — | — | — | 36.2 | 26.3 | 28.2 | 23.2 | 25.8 | |
| OV-3DIS3D mask proposal=3D, evaluation protocol=Top-12026.04 | 31.8 | — | — | — | 36.4 | 27.2 | 23.1 | 22.3 | 24.2 | |
| Open-YOLO 3DInput Modality=P & I & D & C2026.03 | 31.7 | — | — | — | 36.2 | — | — | — | 24.7 | |
| Open-YOLO3D3D mask proposal=3D, evaluation protocol=Top-k2026.04 | 31.7 | — | — | — | 36.2 | 27.8 | 21.6 | 24.3 | 24.7 | |
| PE+TR+AvgPool3D mask proposal=3D, evaluation protocol=Top-k2026.04 | 31 | — | — | — | 37.5 | 23.4 | 24.1 | 18.4 | 21.8 | |
| PE+TR+AvgPool3D mask proposal=3D, evaluation protocol=Top-12026.04 | 31 | — | — | — | 37.5 | 23.4 | 24.1 | 18.4 | 21.8 | |
| Ours (Mask3D)Mask supervision=supervised, Semantic supervision=zero-shot, Model variant=Mask3D2024.11 | 30.3 | — | — | 23.2 | 33.3 | — | — | — | — | |
| Open3DISInputs=3D+2D, 3D Region Proposal Network=Superpoints + ISBNet + Grounded-SAM2025.12 | 29.4 | — | — | — | 32.8 | 27.8 | 21.8 | — | — | |
| Open3DIS (2D and 3D)Setting=Open-vocab, 3D Proposal=ISBNet [50]2023.12 | 29.4 | — | — | 23.7 | 32.8 | 27.8 | 21.8 | 21.2 | — | |
| Open3DISMask supervision=supervised, Semantic supervision=zero-shot2024.11 | 29.4 | — | — | 23.7 | 32.8 | — | — | — | — | |
| Open3DIS3D mask proposal=2D+3D, evaluation protocol=Top-k2026.04 | 29.4 | — | — | — | 32.8 | 27.8 | 21.8 | 21.2 | 23.7 | |
| Open-YOLO3D3D mask proposal=3D, evaluation protocol=Top-12026.04 | 28.3 | — | — | — | 31.7 | 25.6 | 18.5 | 21 | 21.9 | |
| CLIP onlypre-training=CLIP embeddings2022.04 | 27.91 | 64.21 | 27.58 | — | — | — | — | — | — | |
| Open3DIS (only 2D)Setting=Open-vocab, 3D Proposal=None2023.12 | 26.1 | — | — | 18.2 | 31.4 | 18.9 | 19.2 | 16.5 | — | |
| Language-Grounded 3D Pre-trainingbatch size=8, epochs=300, optimizer=momentum SGD, starting learning rate=0.022022.04 | 26.09 | 65.32 | 27.72 | — | — | — | — | — | — | |
| OursMask supervision=zero-shot, Semantic supervision=zero-shot2024.11 | 25.8 | — | — | 14.3 | 33.6 | — | — | — | — | |
| CSCpre-training=CSC2022.04 | 25.24 | 63.13 | 25.92 | — | — | — | — | — | — | |
| OVIR-3DSetting=Open-vocab, 3D Proposal=None2023.12 | 24.9 | — | — | 13 | 32.3 | 14.4 | 11.7 | 12.7 | — | |
| Scratchpre-training=none2022.04 | 24.47 | 61.04 | 25.37 | — | — | — | — | — | — | |
| Mask-ClusteringMask supervision=zero-shot, Semantic supervision=zero-shot2024.11 | 23.3 | — | — | 12 | 30.1 | — | — | — | — | |
| Open3DIS (only 3D)Setting=Open-vocab, 3D Proposal=ISBNet [50]2023.12 | 23.1 | — | — | 18.6 | 27.3 | 24.7 | 13.3 | 16.9 | — | |
| Open3DIS (3D only)Mask supervision=supervised, Semantic supervision=zero-shot, Model variant=3D only2024.11 | 23.1 | — | — | 18.6 | 27.3 | — | — | — | — | |
| OpenMask3DSetting=Open-vocab, 3D Proposal=Mask3D [60]2023.12 | 19.9 | — | — | 15.4 | 23.1 | 17.1 | 14.9 | 14.1 | — | |
| OpenMask3D (Mask3D)Mask supervision=supervised, Semantic supervision=zero-shot, Model variant=Mask3D2024.11 | 19.9 | — | — | 15.4 | 23.1 | — | — | — | — | |
| OpenMask3DInput Modality=P & I & D & C2026.03 | 19.9 | — | — | — | 23.1 | — | — | — | 15.4 | |
| SAI3DInputs=3D+2D, 3D Region Proposal Network=Superpoints + SAM2025.12 | 18.8 | — | — | — | 24.1 | 12.1 | 16.2 | — | — | |
| SAI3DMask supervision=zero-shot, Semantic supervision=zero-shot2024.11 | 18.8 | — | — | 12.7 | 24.1 | — | — | — | — | |
| OVIR-3DMask supervision=zero-shot, Semantic supervision=zero-shot2024.11 | 18.7 | — | — | 9.3 | 25 | — | — | — | — | |
| ChorusInputs=3D, 3D Region Proposal Network=Mask3D2025.12 | 18 | — | — | — | 19.6 | 18.5 | 13 | — | — | |
| Mosaic3DInputs=3D, 3D Region Proposal Network=Mask3D2025.12 | 16 | — | — | — | 17.8 | 21.8 | 5.4 | — | — | |
| OpenScene + Mask3DSetting=Open-vocab, 3D Proposal=Mask3D [60]2023.12 | 15.2 | — | — | 11.7 | 17.8 | 13.4 | 9.9 | 11.6 | — | |
| SAM3DMask supervision=zero-shot, Semantic supervision=zero-shot2024.11 | 15.2 | — | — | 9.8 | 20.7 | — | — | — | — | |
| SAM3DSetting=Open-vocab, 3D Proposal=None2023.12 | 14.2 | — | — | 6.1 | 21.3 | 7 | 4.6 | 6.2 | — | |
| OpenIns3DInputs=3D, 3D Region Proposal Network=Mask3D2025.12 | 10.3 | — | — | — | 14.4 | 16 | 4.2 | — | — | |
| OpenIns3DSetting=Open-vocab, 3D Proposal=Mask3D [60]2023.12 | 10.3 | — | — | 8.8 | 14.4 | 16 | 4.2 | 6.5 | — | |
| RegionPLCInputs=3D, 3D Region Proposal Network=Mask3D2025.12 | 8.6 | — | — | — | 9.7 | 15.6 | 1.7 | — | — | |
| OpenScene + DBScanSetting=Open-vocab, 3D Proposal=None2023.12 | 7.8 | — | — | 2.8 | 18.6 | 2.7 | 2.6 | 3.1 | — | |
| OpenScene-3DInputs=3D, 3D Region Proposal Network=Mask3D2025.12 | 6.2 | — | — | — | 7.2 | 10.6 | 0.7 | — | — | |
| OneFormer3DInput Modality=I, Reconstruction=3D reconstruction with VGGT [52]2026.03 | 4.1 | — | — | — | 6.4 | — | — | — | 2.5 | |
| Any3DIS3D mask proposal=2D+3D, evaluation protocol=Top-k2026.04 | — | — | — | — | — | 27.4 | 26.4 | 23.8 | 25.8 | |
| AvgPool w/ PE+TRevaluation protocol=Top-k2026.04 | — | — | — | — | — | 57.7 | 58 | 42.3 | 52.3 | |
| AvgPool w/ PE+TRevaluation protocol=Top-12026.04 | — | — | — | — | — | 51.9 | 57.9 | 41.4 | 50 | |
| CAMFusionevaluation protocol=Top-k2026.04 | — | — | — | — | — | 70.3 | 65.6 | 51.2 | 62.1 | |
| CAMFusionevaluation protocol=Top-12026.04 | — | — | — | — | — | 60.9 | 61.3 | 46.7 | 56 | |
| OpenYolo3Devaluation protocol=Top-k2026.04 | — | — | — | — | — | 43.3 | 38.5 | 36.8 | 39.6 | |
| OpenYolo3Devaluation protocol=Top-12026.04 | — | — | — | — | — | 43.3 | 38.5 | 36.9 | 39.6 | |
| OV-3DISevaluation protocol=Top-k2026.04 | — | — | — | — | — | 33.9 | 32.5 | 26.5 | 30.9 | |
| OV-3DISevaluation protocol=Top-12026.04 | — | — | — | — | — | 30.3 | 36 | 25.8 | 30.3 | |
| OV-3DIS w/ PE+TRevaluation protocol=Top-k2026.04 | — | — | — | — | — | 51.1 | 50.5 | 38.9 | 46.6 | |
| OV-3DIS w/ PE+TRevaluation protocol=Top-12026.04 | — | — | — | — | — | 43.7 | 54.5 | 38.7 | 45.1 | |
| Search3D3D mask proposal=3D, evaluation protocol=Top-12026.04 | — | — | — | — | — | 26.3 | 21.4 | 21.2 | 23 |