Dichotomous Image Segmentation on DIS ALL 2,000 (test)
91.5FmaxPDFNet
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PDFNetBackbone=Swin-B, Params=94M2025.03 | 91.5 | 87.6 | 94.4 | 91.6 | 0.03 | |
| DiffDISBackbone=SD-Turbo, Params=865M+2025.03 | 91.1 | 89.2 | 95.1 | 90.5 | 0.027 | |
| MVANetBackbone=Swin-B, Params=93M2025.03 | 90.8 | 86.1 | 93.8 | 90.4 | 0.035 | |
| MVANet*Backbone=Swin-B, Params=93M, Input=RGB-depth concatenation2025.03 | 90.6 | 86.2 | 93.5 | 90.7 | 0.033 | |
| BiRefNetBackbone=Swin-L, Params=215M2025.03 | 90 | 85.8 | 93.4 | 90.1 | 0.035 | |
| CPNetBackbone=Swin-B & Swin-B, Params=216M2025.03 | 89.3 | 85.4 | 93.1 | 89.8 | 0.035 | |
| InSPyReNetBackbone=Swin-B, Params=88M2025.03 | 89.1 | 83.8 | 92.2 | 90 | 0.039 | |
| DACODBackbone=ResNet-50 & Swin-B, Params=189M2025.03 | 87.8 | 82.7 | 92.1 | 88.4 | 0.042 | |
| GenPerceptBackbone=SD U-Net, Params=865M+2025.03 | 87.5 | 85.6 | 93.9 | 88.5 | 0.036 | |
| MAGNetBackbone=SMT-T & MbNet-V2, Params=16M2025.03 | 86.9 | 82.3 | 92 | 87.8 | 0.043 | |
| RISNetBackbone=PVT-V2-B2, Params=25M2025.03 | 86.9 | 82.8 | 91.8 | 88 | 0.044 | |
| FP-DIS2025.03 | 83.1 | 77 | 89.5 | 84.7 | 0.057 | |
| UDUNBackbone=ResNet-50, Params=25M2025.03 | 83.1 | 77.2 | 89.2 | 84.4 | 0.057 | |
| IS-NetBackbone=U2-Net, Params=44M2025.03 | 79.9 | 72.5 | 85.8 | 81.9 | 0.07 |