Reconstruction on ShapeNet In-Context
2.2CD L1DeformPIC
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DeformPICTraining Paradigm=In-Context Learning2026.04 | 2.2 | 2.3 | 2.5 | 2.8 | 3.9 | 2.7 | |
| PCTModel training/learning paradigm=Task-specific models (trained separately), Venues=CVM'212024.11 | 2.4 | 2.4 | 2.5 | 2.6 | 3 | 2.6 | |
| ACTModel training/learning paradigm=Task-specific models (trained separately), Venues=ICLR'232024.11 | 2.4 | 2.5 | 2.3 | 2.5 | 2.8 | 2.5 | |
| PCTTraining Strategy=Task-Specific, Task Head=Single task2024.04 | 2.4 | 2.4 | 2.5 | 2.6 | 3 | 2.6 | |
| ACTTraining Strategy=Task-Specific, Task Head=Single task2024.04 | 2.4 | 2.5 | 2.3 | 2.5 | 2.8 | 2.5 | |
| PCTTraining Paradigm=Task-Specific2026.04 | 2.4 | 2.4 | 2.5 | 2.6 | 3 | 2.6 | |
| ACTTraining Paradigm=Task-Specific2026.04 | 2.4 | 2.5 | 2.3 | 2.5 | 2.8 | 2.5 | |
| PIC-CatModel training/learning paradigm=In-context learning models, Venues=NeurIPS'232024.11 | 3.2 | 3.6 | 4.6 | 4.9 | 5.5 | 4.3 | |
| PICTraining Strategy=In-context Learning, Architecture=task-agnostic for multiple tasks, Aggregation=-Cat2024.04 | 3.2 | 3.6 | 4.6 | 4.9 | 5.5 | 4.3 | |
| PIC-CatTraining Paradigm=In-Context Learning2026.04 | 3.2 | 3.6 | 4.6 | 4.9 | 5.5 | 4.3 | |
| PointNetModel training/learning paradigm=Task-specific models (trained separately), Venues=CVPR'172024.11 | 3.7 | 3.7 | 3.8 | 4.1 | 4.3 | 3.9 | |
| PointNetTraining Strategy=Task-Specific, Task Head=Single task2024.04 | 3.7 | 3.7 | 3.8 | 3.9 | 4.1 | 3.9 | |
| PointNetTraining Paradigm=Task-Specific2026.04 | 3.7 | 3.7 | 3.8 | 3.9 | 4.1 | 3.9 | |
| PIC-Sep + MICASModel training/learning paradigm=In-context learning models, Venues=Ours2024.11 | 3.8 | 3.9 | 4 | 4.4 | 5.6 | 4.3 | |
| DGCNNModel training/learning paradigm=Task-specific models (trained separately), Venues=TOG' 192024.11 | 3.9 | 3.9 | 4 | 4.1 | 4.3 | 4 | |
| DGCNNTraining Strategy=Task-Specific, Task Head=Single task2024.04 | 3.9 | 3.9 | 4 | 4.1 | 4.3 | 4 | |
| DGCNNTraining Paradigm=Task-Specific2026.04 | 3.9 | 3.9 | 4 | 4.1 | 4.3 | 4 | |
| UniPre3DTraining Paradigm=Multitask, Pre-trained=Yes2026.04 | 3.9 | 4 | 4.2 | 4.3 | 5.6 | 4.4 | |
| PCP-MAETraining Paradigm=Multitask, Pre-trained=Yes2026.04 | 4 | 4.1 | 4.4 | 4.6 | 6 | 4.6 | |
| PIC++Training Strategy=In-context Learning, Architecture=task-agnostic for multiple tasks, Aggregation=-Cat2024.04 | 4.5 | 3.3 | 3.7 | 4.6 | 5.4 | 4.3 | |
| PIC++Training Strategy=In-context Learning, Architecture=task-agnostic for multiple tasks, Aggregation=-Sep2024.04 | 4.5 | 4.1 | 4.2 | 4.2 | 5.9 | 4.6 | |
| PIC-S-SepModel training/learning paradigm=In-context learning models, Venues=Arxiv'242024.11 | 4.6 | 4.5 | 4.5 | 4.8 | 7.1 | 5.1 | |
| PIC-Cat + MICASModel training/learning paradigm=In-context learning models, Venues=Ours2024.11 | 4.6 | 4.2 | 4.5 | 4.8 | 5.7 | 4.7 | |
| PIC-S-SepTraining Paradigm=In-Context Learning2026.04 | 4.6 | 4.5 | 4.5 | 4.8 | 7.1 | 5.1 | |
| PIC-SepModel training/learning paradigm=In-context learning models, Venues=NeurIPS'232024.11 | 4.7 | 4.3 | 4.3 | 4.4 | 4.7 | 4.5 | |
| PICTraining Strategy=In-context Learning, Architecture=task-agnostic for multiple tasks, Aggregation=-Sep2024.04 | 4.7 | 4.3 | 4.3 | 4.4 | 5.7 | 4.7 | |
| PIC-SepTraining Paradigm=In-Context Learning2026.04 | 4.7 | 4.3 | 4.3 | 4.4 | 5.7 | 4.7 | |
| Point-MAEModel training/learning paradigm=Multi-task models: share backbone + multi-task heads, Venues=ECCV'222024.11 | 5.5 | 5.5 | 6.1 | 6.4 | 6.4 | 6 | |
| PointMAE‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 5.5 | 5.5 | 6.1 | 6.4 | 6.4 | 6 | |
| Point-MAETraining Paradigm=Multitask, Pre-trained=Yes2026.04 | 5.5 | 5.5 | 6.1 | 6.4 | 6.4 | 6 | |
| ACTModel training/learning paradigm=Multi-task models: share backbone + multi-task heads, Venues=ICLR'232024.11 | 6.6 | 6.5 | 6.6 | 7 | 6.8 | 6.7 | |
| ACT‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 7.4 | 6.6 | 6.5 | 6.6 | 7 | 6.8 | |
| ACTTraining Paradigm=Multitask, Pre-trained=Yes2026.04 | 7.4 | 6.6 | 6.5 | 6.6 | 7 | 6.8 | |
| PCMamba‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 8.1 | 8.9 | 10.2 | 10.5 | 11.7 | 9.9 | |
| MAMBA3D‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 8.4 | 7.8 | 9.3 | 9.8 | 10.5 | 9.2 | |
| PIC-S-CatModel training/learning paradigm=In-context learning models, Venues=Arxiv'242024.11 | 9.3 | 5.1 | 4.8 | 5.1 | 5.5 | 5.9 | |
| PIC-S-CatTraining Paradigm=In-Context Learning2026.04 | 9.3 | 5.1 | 4.8 | 5 | 10.3 | 6.9 | |
| ReConModel training/learning paradigm=Multi-task models: share backbone + multi-task heads, Venues=ICML 232024.11 | 12.4 | 12.1 | 12.4 | 12.5 | 13.1 | 12.5 | |
| ReCon‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 12.4 | 12.1 | 12.4 | 12.5 | 13.1 | 12.5 | |
| ReConTraining Paradigm=Multitask, Pre-trained=Yes2026.04 | 12.4 | 12.1 | 12.4 | 12.5 | 13.1 | 12.5 | |
| I2P-MAEModel training/learning paradigm=Multi-task models: share backbone + multi-task heads, Venues=CVPR'232024.11 | 17 | 16 | 16.7 | 17.2 | 18.5 | 17.2 | |
| I2P-MAE‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 17 | 16 | 16.7 | 17.2 | 18.5 | 17.2 | |
| I2P-MAETraining Paradigm=Multitask, Pre-trained=Yes2026.04 | 17 | 16 | 16.7 | 17.2 | 18.5 | 17.2 | |
| PCTModel training/learning paradigm=Multi-task models: share backbone + multi-task heads, Venues=CVM'212024.11 | 34.7 | 44.1 | 49.9 | 50 | 52.3 | 46.2 | |
| PCT†Training Strategy=Multitask, Backbone=Shared backbone, Task Head=Multiple task heads2024.04 | 34.7 | 44.1 | 49.9 | 50 | 52.3 | 46.2 | |
| PCTTraining Paradigm=Multitask, Pre-trained=No2026.04 | 34.7 | 44.1 | 49.9 | 50 | 52.3 | 46.2 | |
| DGCNNModel training/learning paradigm=Multi-task models: share backbone + multi-task heads, Venues=TOG' 192024.11 | 38.8 | 36.6 | 37.5 | 37.9 | 42.9 | 37.7 | |
| DGCNN†Training Strategy=Multitask, Backbone=Shared backbone, Task Head=Multiple task heads2024.04 | 38.8 | 36.6 | 37.5 | 37.9 | 42.9 | 37.7 | |
| DGCNNTraining Paradigm=Multitask, Pre-trained=No2026.04 | 38.8 | 36.6 | 37.5 | 37.9 | 42.9 | 37.7 | |
| DGCNN‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 46.7 | 47.2 | 48.1 | 48.6 | 48.5 | 47.8 | |
| PointNet‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 47 | 45.8 | 45.4 | 45.4 | 45.8 | 45.9 | |
| PCT‡Training Strategy=Multitask, Backbone=Pre-trained shared backbone, Task Head=Multiple task heads2024.04 | 64.7 | 60.8 | 59.2 | 60.1 | 59.7 | 61 | |
| PointNetModel training/learning paradigm=Multi-task models: share backbone + multi-task heads, Venues=CVPR'172024.11 | 87.2 | 86.6 | 87.3 | 90.8 | 92.2 | 88.8 | |
| PointNet†Training Strategy=Multitask, Backbone=Shared backbone, Task Head=Multiple task heads2024.04 | 87.2 | 86.6 | 87.3 | 90.8 | 92.2 | 88.8 | |
| PointNetTraining Paradigm=Multitask, Pre-trained=No2026.04 | 87.2 | 86.6 | 87.3 | 90.8 | 92.2 | 88.8 | |
| CopyModel training/learning paradigm=In-context learning models2024.11 | 155 | 153 | 152 | 156 | 155 | 154 | |
| CopyTraining Strategy=In-context Learning, Architecture=task-agnostic for multiple tasks2024.04 | 155 | 153 | 152 | 156 | 155 | 154 | |
| Point-BERTModel training/learning paradigm=In-context learning models, Venues=CVPR 222024.11 | 288 | 285 | 292 | 286 | 308 | 292 | |
| Point-BERTTraining Strategy=In-context Learning, Architecture=task-agnostic for multiple tasks2024.04 | 288 | 285 | 292 | 286 | 308 | 292 |