Table 2 Performance of the models over the training and validation dataset.
Models | Classes | IoU (%) | Recall (%) | Precision (%) | mIoU (%) | mPA (%) | Accuracy (%) | Params (M) | Time |
|---|---|---|---|---|---|---|---|---|---|
UNet | W B | 92.65 72.02 | 98.22 76.93 | 94.23 91.86 | 82.34 | 87.58 | 93.82 | 97.25 | 8h10min |
DeepLabv3+ | W B | 92.20 69.90 | 98.42 74.14 | 93.59 92.45 | 81.05 | 86.28 | 93.40 | 22.98 | 7h39min |
HRNet | W B | 92.79 72.36 | 98.40 76.80 | 94.21 92.60 | 82.58 | 87.60 | 93.93 | 38.45 | 7h25min |
SegFormer | W B | 93.07 73.68 | 98.26 78.60 | 94.63 92.18 | 83.38 | 88.43 | 94.19 | 14.58 | 5h28min |
SegFormer (RGB + TIFF) | W B | 92.88 72.90 | 98.27 77.73 | 94.42 92.15 | 82.89 | 88.0 | 94.02 | 14.60 | 5h35min |
Tiff-SegFormer | W B | 93.48 75.08 | 98.48 79.45 | 94.84 93.18 | 84.28 | 88.97 | 94.55 | 28.63 | 5h48min |
