OralBBNet: UFBA-425 dataset analysis

[OralBBNet: Spatially Guided Dental Segmentation of Panoramic X-Rays with Bounding Box Priors]

Paper Link:https://arxiv.org/pdf/2406.03747

 

[UFBA-425]

Dataset Link: https://figshare.com/articles/dataset/UFBA-425/29827475

 

[Dataset Overview]

• 425 panoramic dental X-ray images
  • Bounding box annotations
  • Instance segmentation masks
  • Structured metadata (metadata.csv)
• Within folder
  • numbering_xrays/ : panoramic X-ray images used for bounding box detection
  • segmentation_xrays/ : X-ray images for segmentation processing
  • bounding_boxes/ : YOLO-format text files with tooth-level bounding box annotations
    • Includes: Class ID (0-31), Normalized center coordinates (x,y), Normalized width and height
    • Class ID (0-31) = FDI (11-18, 21-28, 31-38, 41-48) → total 32
  • polygon_masks/ : Per-tooth instance segmentation masks stored as .ome.tiff files
  • metadata.csv : Converts the bounding_boxes/ info into pixel-based coordinates: (Bounding box format)
    • FDI_XX_x – top-left x coordinate
    • FDI_XX_y – top-left y coordinate
    • FDI_XX_w – bounding box width
    • FDI_XX_h – bounding box height
  • README.md : Explains dataset usage and structure

 

[bounding_box file]

• Total 425 .txt files
• One text file has 32 lines (32 detected teeth). Each line describes position of one tooth
  • class_id x_center y_center width height
    • class_id: 0-31 using FDI tooth numbers
    • x_center: horizontal center, normalized (0~1, 0: left, 1: right)
    • y_center: vertical center, normalized (0~1, 0: top, 1: bottom)
    • width
    • height

[Numbering_xrays file → object detection]

• Total 425 jpg files, 512 x 512 (1:1 correspondence with bounding_box .txt file)
  • For example,
    • cate1-00002_jpg.rf.e0c956c6468ef96b53c862916e6fb6e8.txt
    • cate1-00002_jpg.rf.e0c956c6468ef96b53c862916e6fb6e8.jpg

[Segmentation_xrays]

• Total 425 jpg files

[Polygon_masks]

• File name: cate1-00005_42.ome.tiff
  • _FDI (11-18, 21-28, 31-38, 41-48) number → total 32
  • _background → 1

[Metadata.xlsx]

metadata_edited.xlsx

0.37MB

• Bounding box
  • Missing Teeth (BBox Missing): 2025, 2025/1600 = 14.89%
• Segmentation Masks
  • Missing masks: 2030, 2030/13600 = 14.93%

 

 

[Visualization]

Bounding box, Numbering_xrays 활용 → cate1-00002

 

segmentation_xrays: cate1-00002 polygon mask: fdi 11

[Github]

https://github.com/devichand579/Instance_seg_teeth

GitHub - devichand579/Instance_seg_teeth: UFBA-425 Dataset and code for OralBBNet