| Title |
Semantic Segmentation of Teeth using Layered UNet |
| Authors |
김태훈(Tae-Hoon Kim) ; 박종진(Jong-Jin Park) |
| DOI |
https://doi.org/10.5370/KIEE.2023.72.11.1470 |
| Keywords |
Teeth;Semantic segmentation; Deep learning; Layered UNet; AI model |
| Abstract |
In this paper, using teeth data set provided by AIHub individual tooth semantic segmentation was carried out through layered UNet.
AIHub's dental data set is provided as 2D panoramic X-ray images and 3D CBCT images to develop dental AI models. The layered UNet proposed in the previous paper showed that it also learned excellently on teeth data sets that are distinguished for each tooth by assigning tooth numbers.
As a result of the simulation, the learning results by the layered UNet model showed loss function values of 0.005 and 0.006 for training and validation data, respectively. Accuracy and IoU used as other evaluation indicators showed results of (0.99, 0.99) and (0.89, 0.88) for （training and validation data）, respectively. the learned layered UNet was applied to the existing tooth data to segment the teeth and stack the results for each slice to extract the 3D tooth shape. Although there are some parts where the learning was not done well, the teeth were extracted being well distinguished by tooth, and it was shown that individual tooth could also be extracted. |