Table 3 Performance evaluation of various segmentation models by training on a different number of randomly sampled sets selected from the actual training set of our dataset for supervised learning.
From: Dental caries detection using a semi-supervised learning approach
Model | Backbone | No. of Labelled Samples | ||||
|---|---|---|---|---|---|---|
20 | 40 | 60 | 80 | 120 | ||
mIoU | ||||||
FCN | ResNet-101 | 44.13 | 48.87 | 48.16 | 47.91 | 47.83 |
PSPNet | ResNet-101 | 42.84 | 45.91 | 45.29 | 44.50 | 44.28 |
LRASPP | MobileNet-v3 | 43.47 | 47.38 | 46.27 | 45.86 | 45.65 |
FPN | ResNet-101 | 44.21 | 48.65 | 47.95 | 47.33 | 47.06 |
LinkNet | ResNet-101 | 43.46 | 47.87 | 47.54 | 47.02 | 46.87 |
Deeplab-v3 | ResNet-101 | 46.57 | 52.41 | 51.63 | 50.47 | 50.18 |
