Binary Segmentation on Kvasir-SEG (test)
0.9355DSCt-vMF Dice loss
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| t-vMF Dice lossArchitecture=TransUNet, k=22022.07 | 0.9355 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossArchitecture=TransUNet, lambda=22022.07 | 0.9355 | — | — | — | — | — | |
| Noise-robust Dice lossArchitecture=TransUNet2022.07 | 0.9353 | — | — | — | — | — | |
| Dice lossArchitecture=TransUNet2022.07 | 0.935 | — | — | — | — | — | |
| Focal Dice lossArchitecture=TransUNet2022.07 | 0.935 | — | — | — | — | — | |
| Focal Tversky lossArchitecture=TransUNet2022.07 | 0.9342 | — | — | — | — | — | |
| t-vMF Dice lossArchitecture=TransUNet, k=322022.07 | 0.9339 | — | — | — | — | — | |
| t-vMF Dice lossArchitecture=TransUNet, k=1282022.07 | 0.9339 | — | — | — | — | — | |
| Generalised Dice lossArchitecture=TransUNet2022.07 | 0.9321 | — | — | — | — | — | |
| PVT-EMCAD-B2#Params=26.76M, #FLOPs=5.6G, Resolution=256x2562024.05 | 0.9275 | — | — | — | — | — | |
| WS Dice lossArchitecture=TransUNet2022.07 | 0.9267 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossArchitecture=TransUNet, lambda=322022.07 | 0.9266 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossArchitecture=TransUNet, lambda=1282022.07 | 0.926 | — | — | — | — | — | |
| t-vMF Dice lossArchitecture=U-Net, k=1282022.07 | 0.9245 | — | — | — | — | — | |
| t-vMF Dice lossArchitecture=U-Net, k=322022.07 | 0.9231 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossArchitecture=U-Net, lambda=1282022.07 | 0.9224 | — | — | — | — | — | |
| BFL-log(Dice) lossArchitecture=TransUNet2022.07 | 0.9222 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossArchitecture=U-Net, lambda=322022.07 | 0.9218 | — | — | — | — | — | |
| PVT-CASCADE#Params=34.12M, #FLOPs=7.62G, Resolution=256x2562024.05 | 0.9205 | — | — | — | — | — | |
| PVT-EMCAD-B0#Params=3.92M, #FLOPs=0.84G, Resolution=256x2562024.05 | 0.9195 | — | — | — | — | — | |
| BFL-log(Dice) lossArchitecture=U-Net2022.07 | 0.918 | — | — | — | — | — | |
| PolypPVT#Params=25.11M, #FLOPs=5.30G, Resolution=256x2562024.05 | 0.9156 | — | — | — | — | — | |
| SSFormer-L#Params=66.22M, #FLOPs=17.28G, Resolution=256x2562024.05 | 0.9147 | — | — | — | — | — | |
| TransUNet#Params=105.32M, #FLOPs=38.52G, Resolution=256x2562024.05 | 0.9108 | — | — | — | — | — | |
| t-vMF Dice lossArchitecture=U-Net, k=22022.07 | 0.9101 | — | — | — | — | — | |
| Focal Dice lossArchitecture=U-Net2022.07 | 0.9098 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossArchitecture=U-Net, lambda=22022.07 | 0.9066 | — | — | — | — | — | |
| Focal Tversky lossArchitecture=U-Net2022.07 | 0.9061 | — | — | — | — | — | |
| Dice lossArchitecture=U-Net2022.07 | 0.9034 | — | — | — | — | — | |
| Noise-robust Dice lossArchitecture=U-Net2022.07 | 0.9034 | — | — | — | — | — | |
| WS Dice lossArchitecture=U-Net2022.07 | 0.903 | — | — | — | — | — | |
| TransFuse#Params=143.74M, #FLOPs=82.71G, Resolution=256x2562024.05 | 0.9024 | — | — | — | — | — | |
| UACANet-L#Params=69.16M, #FLOPs=31.51G, Resolution=256x2562024.05 | 0.9017 | — | — | — | — | — | |
| Telescopic AdaptersEvaluation Protocol=Adapter Fine-Tuning, Parameters=613k2025.12 | 0.8979 | 0.835 | — | — | — | — | |
| CaraNet#Params=46.64M, #FLOPs=11.48G, Resolution=256x2562024.05 | 0.8974 | — | — | — | — | — | |
| CRISEvaluation Protocol=End-to-End fine-tuning, Parameters=147M2025.12 | 0.8943 | 0.8337 | — | — | — | — | |
| DeepLabv3+#Params=39.76M, #FLOPs=14.92G, Resolution=256x2562024.05 | 0.8906 | — | — | — | — | — | |
| CLIPSeg DA VLCEvaluation Protocol=Adapter Fine-Tuning, Parameters=3.2M2025.12 | 0.8906 | 0.8228 | — | — | — | — | |
| Generalised Dice lossArchitecture=U-Net2022.07 | 0.887 | — | — | — | — | — | |
| nnU-NetSupervision=Full2025.05 | 0.8841 | — | — | — | — | 0.0047 | |
| CLIPSegEvaluation Protocol=End-to-End fine-tuning, Parameters=150M2025.12 | 0.8769 | 0.8172 | — | — | — | — | |
| SwinUNet#Params=27.17M, #FLOPs=6.2G, Resolution=224x2242024.05 | 0.8761 | — | — | — | — | — | |
| CLIPSeg SA VLCEvaluation Protocol=Adapter Fine-Tuning, Parameters=4.2M2025.12 | 0.868 | 0.7933 | — | — | — | — | |
| I-MedSAMEvaluation Protocol=LoRA Fine-Tuning, Parameters=1.6M2025.12 | 0.8636 | 0.7866 | — | — | — | — | |
| PraNet#Params=32.55M, #FLOPs=6.93G, Resolution=256x2562024.05 | 0.8482 | — | — | — | — | — | |
| AttnUNet#Params=34.88M, #FLOPs=66.64G, Resolution=256x2562024.05 | 0.8349 | — | — | — | — | — | |
| UNet++#Params=9.16M, #FLOPs=34.65G, Resolution=256x2562024.05 | 0.8336 | — | — | — | — | — | |
| UNet#Params=24.53M, #FLOPs=65.53G, Resolution=256x2562024.05 | 0.8287 | — | — | — | — | — | |
| U-NetSupervision=Full2025.05 | 0.8212 | — | — | — | — | 0.0111 | |
| ResUNet++Input Resolution=256x256, Optimizer=Adam, Epochs=1202019.11 | 0.8133 | 0.7927 | 70.64 | 87.74 | — | — | |
| GradTrackSupervision=Gaze2025.05 | 0.8101 | — | — | — | — | 0.0066 | |
| ResUNet-modInput Resolution=256x256, Loss Function=Dice coefficient loss, Epochs=1202019.11 | 0.7909 | 0.4287 | 69.09 | 87.13 | — | — | |
| UNeXt#Params=1.47M, #FLOPs=0.57G, Resolution=256x2562024.05 | 0.7788 | — | — | — | — | — | |
| GazeMedSegSupervision=Gaze2025.05 | 0.778 | — | — | — | — | 0.0102 | |
| AGMMSupervision=Point2025.05 | 0.7557 | — | — | — | — | 0.0084 | |
| BoxTeacherSupervision=Box2025.05 | 0.7333 | — | — | — | — | 0.013 | |
| VAMD-CRFSupervision=Gaze2025.05 | 0.7312 | — | — | — | — | 0.006 | |
| PointSupSupervision=Point2025.05 | 0.7305 | — | — | — | — | 0.0164 | |
| VAMSupervision=Gaze2025.05 | 0.7221 | — | — | — | — | 0.0071 | |
| U-NetInput Resolution=256x256, Epochs=1202019.11 | 0.7147 | 0.4334 | 63.06 | 92.22 | — | — | |
| SANEvaluation Protocol=Adapter Fine-Tuning, Parameters=8.4M2025.12 | 0.6958 | 0.5805 | — | — | — | — | |
| DMPLSSupervision=Scribble2025.05 | 0.6923 | — | — | — | — | 0.0032 | |
| AGMMSupervision=Scribble2025.05 | 0.6723 | — | — | — | — | 0.0102 | |
| USTMSupervision=Scribble2025.05 | 0.6631 | — | — | — | — | 0.0093 | |
| BoxInstSupervision=Box2025.05 | 0.6572 | — | — | — | — | 0.0297 | |
| CLIPSegEvaluation Protocol=Zero-Shot, Parameters=150M2025.12 | 0.5342 | 0.4248 | — | — | — | — | |
| ResUNetInput Resolution=256x256, Loss Function=MSE, Epochs=1202019.11 | 0.5144 | 0.4364 | 50.41 | 72.92 | — | — | |
| ResUnetPRN post-processing=false2022.11 | — | 0.468 | — | — | 45.7 | — | |
| ResUnet + PRNPRN post-processing=true2022.11 | — | 0.529 | — | — | 52.5 | — | |
| ResUnet++PRN post-processing=false2022.11 | — | 0.559 | — | — | 56.8 | — | |
| ResUnet++ + PRNPRN post-processing=true2022.11 | — | 0.617 | — | — | 62.9 | — | |
| SSFormer-SPRN post-processing=false2022.11 | — | 0.868 | — | — | 69.7 | — | |
| SSFormer-S + PRNPRN post-processing=true2022.11 | — | 0.891 | — | — | 72.3 | — | |
| U-NetPRN post-processing=false2022.11 | — | 0.415 | — | — | 38.8 | — | |
| U-Net + PRNPRN post-processing=true2022.11 | — | 0.478 | — | — | 46.3 | — |