Table 2 Cross-domain performance evaluation of HCIA compared to other SOTA models. Optimal results are highlighted in bold.
From: Hierarchical contextual information aggregation for polyp segmentation
Methods | Kvasir \(\rightarrow\) CVC | CVC \(\rightarrow\) Kvasir | ||
|---|---|---|---|---|
mDice | mIoU | mDice | mIoU | |
U-Net4 | 0.589 | 0.501 | 0.522 | 0.401 |
ResUNet32 | 0.333 | 0.236 | 0.328 | 0.217 |
ResUNet++6 | 0.560 | 0.470 | 0.302 | 0.204 |
Deeplabv3+54 | 0.650 | 0.538 | 0.674 | 0.532 |
PraNet10 | 0.784 | 0.707 | 0.791 | 0.703 |
MSRF-Net55 | 0.792 | 0.649 | 0.757 | 0.633 |
DCRNet40 | 0.724 | 0.630 | 0.727 | 0.631 |
Transfuse19 | 0.772 | 0.673 | 0.768 | 0.669 |
SSFormer-S49 | 0.796 | 0.722 | 0.779 | 0.697 |
SSFormer-L49 | 0.833 | 0.757 | 0.827 | 0.734 |
FCBFormer52 | 0.873 | 0.803 | 0.884 | 0.821 |
HCIA(ours) | 0.873 | 0.801 | 0.885 | 0.824 |
