Robotic Image Segmentation on OCID
94.1mIoUFull FT (SAM 2-L)
Evaluation Results
| Method | Links | |
|---|---|---|
| Full FT (SAM 2-L)Params=224M2026.05 | 94.1 | |
| Full FT (SAM ViT-H)Params=632M2026.05 | 93.2 | |
| RepSAMParams=4.0M2026.05 | 91.8 | |
| SupervisedPrompt type=box2023.12 | 91.24 | |
| SupervisedPrompt type=point2023.12 | 89.22 | |
| WeSAMPrompt type=box2023.12 | 88.09 | |
| SHOTPrompt type=box2023.12 | 88.06 | |
| TENTPrompt type=box2023.12 | 87.77 | |
| WDASSPrompt type=box2023.12 | 87.68 | |
| TRIBEPrompt type=box2023.12 | 86.77 | |
| DirectPrompt type=box2023.12 | 86.35 | |
| Soft TeacherPrompt type=box2023.12 | 84.98 | |
| DoRAParams=4.1M2026.05 | 82.3 | |
| HQ-SAMParams=610M2026.05 | 82.1 | |
| DePTPrompt type=box2023.12 | 82 | |
| VeRAParams=0.9M2026.05 | 81.9 | |
| WeSAMPrompt type=point2023.12 | 80.14 | |
| AdaLoRAParams=4.5M2026.05 | 80.1 | |
| SupervisedPrompt type=poly2023.12 | 79.23 | |
| TENTPrompt type=poly2023.12 | 77.53 | |
| WeSAMPrompt type=poly2023.12 | 77.41 | |
| WDASSPrompt type=point2023.12 | 77.13 | |
| WDASSPrompt type=poly2023.12 | 76.7 | |
| TRIBEPrompt type=poly2023.12 | 76.5 | |
| SHOTPrompt type=poly2023.12 | 76.25 | |
| LoRA (uniform r=8)Params=4.2M2026.05 | 75.8 | |
| SHOTPrompt type=point2023.12 | 74.39 | |
| SAM-DSAParams=5.1M2026.05 | 74.3 | |
| FastSAMParams=68M2026.05 | 73.8 | |
| Soft TeacherPrompt type=poly2023.12 | 73.75 | |
| DirectPrompt type=poly2023.12 | 72.81 | |
| DirectPrompt type=point2023.12 | 71.41 | |
| DePTPrompt type=poly2023.12 | 70.92 | |
| SAM-AdapterParams=4.8M2026.05 | 70.5 | |
| Soft TeacherPrompt type=point2023.12 | 68.46 | |
| TRIBEPrompt type=point2023.12 | 67.86 | |
| TENTPrompt type=point2023.12 | 66.61 | |
| EfficientSAMParams=13M2026.05 | 65.2 | |
| MobileSAMParams=10M2026.05 | 62.8 | |
| DePTPrompt type=point2023.12 | 56.52 | |
| SAM 2-L (zero-shot)Params=224M2026.05 | 55.1 | |
| SAM ViT-H (zero-shot)Params=—2026.05 | 51.3 |