Table 4 Comparison of quantitative results of different methods on the DRISHTI-GS dataset. Some of the results are derived from [8]
From: Self-supervised pre-training for joint optic disc and cup segmentation via attention-aware network
Method | Year | Params (M) | FLOPs (G) | Times (ms) | OD | OC | ||
|---|---|---|---|---|---|---|---|---|
F1 | BLE | F1 | BLE | |||||
FCNÂ [32] | 2014 | 48.2 | 136.2 | 500 | 0.9321 | 8.90 | 0.8170 | 21.83 |
U-Net [12] | 2015 | 65.9 | 158.3 | 623 | 0.9600 | 7.23 | 0.8500 | 19.53 |
M-Net [13] | 2018 | 71.8 | 164.5 | 650 | 0.9590 | 7.97 | 0.866 | 17.05 |
Stack-U-Net [86] | 2018 | 86.6 | 178.0 | 700 | 0.9700 | 6.47 | 0.8900 | 14.39 |
POSALÂ [17] | 2019 | 68.5 | - | - | 0.9650 | - | 0.8580 | - |
CE-Net [85] | 2019 | 71.3 | 125.0 | 550 | 0.9688 | 5.04 | 0.8699 | 16.06 |
JointRCNNÂ [14] | 2020 | - | - | - | 0.9640 | - | 0.8640 | - |
BGA-Net [87] | 2021 | 75.6 | 148.5 | 600 | 0.9750 | 7.01 | 0.8980 | 14.37 |
DCGANÂ [88] | 2022 | 102.9 | - | - | 0.9746 | 7.35 | 0.8631 | 18.69 |
RSAP-Net [8] | 2022 | 87.6 | 235.0 | 800 | 0.9752 | 6.33 | 0.9012 | 11.97 |
Ours | 2023 | 54.6 | 147.6 | 560 | 0.9801 | 6.21 | 0.9087 | 10.07 |