Table 4.
Comparison of Model Performance on ADE20K Validation Set.
| Model | Train set | Prompt | Crop Size | mIoU |
|---|---|---|---|---|
| HorNet (Rao et al., 2022) | ADE20K train set | No | 6402 | 0.575 |
| SeMask (Jain et al., 2023) | ADE20K train set | No | 6402 | 0.570 |
| SwinV2-G (Liu et al., 2022) | ADE20K train set | No | 8962 | 0.593 |
| ViT-Adapter (Chen et al., 2022) | ADE20K train set | No | 8962 | 0.594 |
| Mask DINO (Li et al., 2023a) | ADE20K train set | No | – | 0.595 |
| BEIT-3 (Wang et al., 2023b) | ADE20K train set | No | 8962 | 0.620 |
| SAM2_b (Ravi et al., 2024) | SA-V | MBB | 10242 | 0.757 |
| SAM2_l (Ravi et al., 2024) | SA-V | MBB | 10242 | 0.756 |