Instance Segmentation on ScanNet++ (val)
0.223mAPSonata (full)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Sonata (full)Params=124.8M, Learn.=124.8M, Pct.=100%, Evaluation Protocol=full fine-tuning2025.03 | 0.223 | 0.403 | 0.336 | |
| SonataReal Pretraining Data=18k, Synth Pretraining Data=121k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.223 | — | — | |
| LAM3C*Real Pretraining Data=15k, VGPC Pretraining Data=49k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Large), Pre-training steps=434k2025.12 | 0.223 | — | — | |
| PPT (sup.)Params=124.8M, Learn.=124.8M, Pct.=100%, Evaluation Protocol=full fine-tuning2025.03 | 0.219 | 0.393 | 0.328 | |
| PPTReal Pretraining Data=1k, Synth Pretraining Data=21k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.219 | — | — | |
| MSC (full)Params=124.8M, Learn.=124.8M, Pct.=100%, Evaluation Protocol=full fine-tuning2025.03 | 0.217 | 0.389 | 0.309 | |
| MSCReal Pretraining Data=7k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.217 | — | — | |
| LAM3C*VGPC Pretraining Data=49k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Large), Pre-training steps=434k2025.12 | 0.211 | — | — | |
| PointGroupParams=124.8M, Learn.=124.8M, Pct.=100%2025.03 | 0.209 | 0.367 | 0.307 | |
| LAM3CVGPC Pretraining Data=49k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.205 | — | — | |
| Sonata (all real)Real Pretraining Data=15k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.197 | — | — | |
| LAM3CVGPC Pretraining Data=16k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.197 | — | — | |
| Sonata (dec.)Params=124.8M, Learn.=16.3M, Pct.=13%, Evaluation Protocol=decoder probing2025.03 | 0.188 | 0.381 | 0.291 | |
| Sonata (ScanNet)Real Pretraining Data=1k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.187 | — | — | |
| PTv3Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 0.185 | — | — | |
| LAM3C*Real Pretraining Data=15k, VGPC Pretraining Data=49k, Evaluation Protocol=LP, Backbone=PTv3 (Large), Pre-training steps=434k2025.12 | 0.127 | — | — | |
| Sonata (lin.)Params=124.8M, Learn.=<0.2M, Pct.=<0.2%, Evaluation Protocol=linear probing2025.03 | 0.122 | 0.316 | 0.224 | |
| LAM3C*VGPC Pretraining Data=49k, Evaluation Protocol=LP, Backbone=PTv3 (Large), Pre-training steps=434k2025.12 | 0.121 | — | — | |
| LAM3CVGPC Pretraining Data=49k, Evaluation Protocol=LP, Backbone=PTv3 (Base)2025.12 | 0.113 | — | — | |
| Sonata (all real)Real Pretraining Data=15k, Evaluation Protocol=LP, Backbone=PTv3 (Base)2025.12 | 0.1 | — | — | |
| Sonata (ScanNet)Real Pretraining Data=1k, Evaluation Protocol=LP, Backbone=PTv3 (Base)2025.12 | 0.097 | — | — | |
| LAM3CVGPC Pretraining Data=16k, Evaluation Protocol=LP, Backbone=PTv3 (Base)2025.12 | 0.095 | — | — | |
| MSC (lin.)Params=124.8M, Learn.=<0.2M, Pct.=<0.2%, Evaluation Protocol=linear probing2025.03 | 0.013 | 0.048 | 0.026 | |
| PTv3Evaluation Protocol=LP, Backbone=PTv3 (Base)2025.12 | 0.0003 | — | — | |
| ChorusEvaluation Protocol=linear probing2025.12 | — | 0.37 | 0.279 | |
| ChorusEvaluation Protocol=decoder probing2025.12 | — | 0.419 | 0.338 | |
| ChorusEvaluation Protocol=fine-tuning2025.12 | — | 0.429 | 0.372 | |
| MSCEvaluation Protocol=linear probing2025.12 | — | 0.048 | 0.026 | |
| PTv3Evaluation Protocol=supervised2025.12 | — | 0.414 | 0.325 | |
| SonataEvaluation Protocol=linear probing2025.12 | — | 0.335 | 0.245 | |
| SonataEvaluation Protocol=decoder probing2025.12 | — | 0.394 | 0.335 | |
| SonataEvaluation Protocol=fine-tuning2025.12 | — | 0.41 | 0.353 |