STS-2D-Tooth

The 2D panoramic dental X-ray subset of the STS (Semi-supervised Teeth Segmentation) multi-modal dataset, as released in Wang et al., Scientific Data 12, 117 (2025) and used in the MICCAI 2023 STS Challenge.

Composition

4,000 panoramic X-ray images (PNG, 640x320, 3-channel grayscale-as-RGB) split across two demographic subsets:

Subset	Total	Labeled	Unlabeled
A-PXI (adult)	3,500	850	2,650
C-PXI (child)	500	50	450
Total	4,000	900	3,100

Masks are 1-bit binary tooth-region masks at the same resolution as the source image. Annotations were initialized manually on a 300-image seed by 20 trained dental practitioners, refined by an R2 U-Net assistant model, and quality-vetted by 6 dentists.

Splits

• a_pxi_labeled (850) - adult panoramic X-rays with binary tooth masks
• a_pxi_unlabeled (2,650) - adult panoramic X-rays, no masks
• c_pxi_labeled (50) - paediatric panoramic X-rays with binary tooth masks
• c_pxi_unlabeled (450) - paediatric panoramic X-rays, no masks

For unlabeled splits the mask column is null.

Citation

@article{wang2025sts,
  title   = {A multi-modal dental dataset for semi-supervised deep learning image segmentation},
  author  = {Wang, Yaqi and others},
  journal = {Scientific Data},
  volume  = {12},
  pages   = {117},
  year    = {2025},
  doi     = {10.1038/s41597-024-04306-9}
}

@article{wang2024stschallenge,
  title   = {STS MICCAI 2023 Challenge: Grand challenge on 2D and 3D semi-supervised tooth segmentation},
  author  = {Wang, Yaqi and others},
  journal = {arXiv:2407.13246},
  year    = {2024}
}

License

CC-BY-4.0 (per the Zenodo release at zenodo.org/records/10597292).