Table 5 Efficiency evaluation ofHCIA compared to other SOTA models. Optimal results are highlighted in bold.
From: Hierarchical contextual information aggregation for polyp segmentation
Methods | Year | Type | mDice | mIoU | Params | FLOPs | FPS |
|---|---|---|---|---|---|---|---|
U-Net | 2015 | CNN | 0.815 | 0.742 | 31.0 | 103.5 | 54 |
U-Net++ | 2018 | CNN | 0.817 | 0.740 | 47.2 | 377.5 | 69 |
PraNet | 2020 | CNN | 0.859 | 0.844 | 30.5 | 13.2 | 53 |
Polyp-PVT | 2021 | Transformer | 0.917 | 0.862 | 25.1 | 11.2 | 53 |
Transfuse | 2021 | Transformer | 0.920 | 0.871 | 26.2 | 21.8 | - |
SSFormer | 2022 | Transformer | 0.935 | 0.890 | 29.3 | 19.1 | - |
APCNet | 2023 | CNN | 0.913 | 0.859 | 33.1 | 16.3 | - |
SRaNet | 2023 | CNN | 0.921 | 0.870 | 24.9 | - | - |
MGCBFormer | 2023 | Transformer | 0.933 | 0.887 | 103.4 | 91.1 | - |
PVT-CASCADE | 2023 | Transformer | 0.924 | 0.875 | 35.3 | 15.4 | - |
CTNet | 2024 | Transformer | 0.917 | 0.863 | 44.2 | 15.2 | - |
UM-Net | 2025 | CNN | 0.930 | 0.882 | 22.8 | 15.6 | 50 |
CTHP | 2024 | Transformer | 0.939 | 0.891 | 47.1 | 54.2 | - |
Polyp-LVT | 2024 | Transformer | 0.909 | 0.851 | 25.1 | 13.2 | - |
CAFE-Net | 2024 | Transformer | 0.933 | 0.889 | 35.5 | 16.1 | - |
Polyp-Mamba | 2025 | CNN | 0.919 | 0.867 | 49.5 | 27.9 | - |
EFA-Net | 2025 | CNN | 0.914 | 0.861 | 27.4 | 33.2 | - |
WBANet | 2025 | Transformer | 0.933 | 0.889 | 38.5 | 11.8 | - |
HCIA(ours) | - | Transformer | 0.942 | 0.894 | 25.8 | 13.7 | 51 |
