Robotic Manipulation on UMI Multi-task Robotic Manipulation Dataset
90Throw into trashDP-T-Di (3 experts)
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| DP-T-Di (3 experts)Action Model Architecture=Transformer, Diverse skill learning mechanism=True, Number of experts=32026.01 | 90 | 90 | 50 | 60 | 70 | 90 | 80 | 20 | 80 | 70 | |
| DP-C-Di (5 experts)Action Model Architecture=CNN, Diverse skill learning mechanism=True, Number of experts=52026.01 | 60 | 90 | 80 | 100 | 80 | 100 | 100 | 80 | 80 | 86 | |
| DP-TAction Model Architecture=Transformer, Diverse skill learning mechanism=False2026.01 | 50 | 50 | 0 | 40 | 10 | 40 | 60 | 10 | 50 | 34 | |
| DP-CAction Model Architecture=CNN, Diverse skill learning mechanism=False2026.01 | 40 | 100 | 40 | 30 | 30 | 90 | 90 | 10 | 40 | 52 |