Bits2Bites Dataset

This page describes the Bits2Bites dataset, a comprehensive collection of intra-oral scans designed for dental occlusion analysis and classification. The dataset was developed by the University of Ferrara in collaboration with the University of Modena and Reggio Emilia and comprises 200 pairs of registered intra-oral scans in STL format with detailed occlusion annotations. The dataset provides high-resolution meshes for both upper and lower dental arches, spatially aligned to preserve true occlusal relationships and includes comprehensive multi-label annotations for clinically relevant occlusion classifications.

Features

The Bits2Bites dataset contains 200 pairs of registered intra-oral scans (STL format) of upper and lower dental arches, aligned in a standardized RAS (Right-Anterior-Superior) coordinate system to preserve occlusal relationships. Scans were acquired using Carestream and 3Shape TRIOS scanners to ensure technological variability, with no selection bias to reflect real-world clinical diversity. Each mesh has ~92,201 ± 28,140 vertices and ~182,444 ± 55,862 faces, with bounding-box dimensions of ~65.9 x 53.84 x 17.9mm (width x depth x height).

Key Features of Bits2Bites:

• 🦷 200 pairs of registered intra-oral scans in STL format;
• 📏 High-resolution meshes with detailed geometric specifications;
• 🎯 Multi-label occlusion annotations across sagittal, vertical, and transverse dimensions;
• 📊 Comprehensive clinical classifications including Angle's classification system;
• 🔬 Dual-scanner acquisition for technology variability representation.

Annotations

Annotations were performed by a single orthodontic specialist with five years of experience in the field. Each scan pair includes detailed, clinically relevant occlusion labels across multiple dimensions. Sagittal classifications are provided separately for the left and right sides, following a subset of Angle's standard classification (i.e., Class I, Class II edge-to-edge, Class II full, Class III). Vertical anterior-posterior relationships are labeled as Normal, Deep Bite, or Open Bite. Transverse relationships are recorded as Normal, Cross Bite, or Scissor Bite, using reference teeth. Finally, midline alignment is annotated as Centered or Deviated. This comprehensive, multi-label annotation scheme enables clinically meaningful classification across sagittal, vertical, and transverse dimensions.