Multi-Task Learning Overall Improvement on NYUD v2 (test)
5.8ΔSD (%)Ours (xTAM)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Ours (xTAM)MTL Setup (Task Subset)=S-D-N2022.06 | 5.8 | 6.49 | — | |
| 3-way SPAD-NetMTL Setup (Task Subset)=S-D-N2022.06 | 4.16 | 3.93 | — | |
| Ours (xTAM)MTL Setup (Task Subset)=S-D2022.06 | 3.54 | — | — | |
| 3-way SPAD-NetMTL Setup (Task Subset)=S-D2022.06 | 2.9 | — | — | |
| MTLMTL Setup (Task Subset)=S-D-N2022.06 | 1.89 | 1.76 | — | |
| MTLMTL Setup (Task Subset)=S-D2022.06 | 1.63 | — | — | |
| PAD-NetMTL Setup (Task Subset)=S-D-N2022.06 | -4.32 | -2.97 | — | |
| PAD-NetMTL Setup (Task Subset)=S-D2022.06 | -5.36 | — | — | |
| 3-way SPAD-NetMTL Setup (Task Subset)=S-D-N-E2022.06 | — | 4.14 | 5.27 | |
| MTLMTL Setup (Task Subset)=S-D-N-E2022.06 | — | 1.88 | 1.5 | |
| Ours (xTAM)MTL Setup (Task Subset)=S-D-N-E2022.06 | — | 7.52 | 8.47 | |
| PAD-NetMTL Setup (Task Subset)=S-D-N-E2022.06 | — | -2.92 | -2.24 |