Transformer-Based Tooth Alignment Prediction with Occlusion and Collision Constraints
Zhenxing Dong, Jiazhou Chen
摘要
TeethAlign3D Dataset Description Document Dataset Overview TeethAlign3D is a 3D dental dataset for orthodontic treatment planning, containing 3D dental models of multiple cases and the tooth alignment target matrices. This dataset is mainly used for tooth alignment optimization and visual analysis in orthodontic treatment. Application Requirements Applicants must fully complete and sign the Data Use Agreement (DUA), which can be obtained from the officially designated document "DUA.pdf"; The executed agreement must be submitted via formal email to the designated address: cjz@zjut.edu.cn; The application email must include the following mandatory information:(1) Full name of the applicant and Zenodo account name;(2) Full institutional affiliation and position;(3) A detailed explanation of the intended use of the dataset. Formal Instructions for Dataset Access Request After submitting your application via email, please proceed to submit a "Request Access" through the dataset webpage. ⚠ Important Notice:You must use the Zenodo account associated with the email domain from which the DUA was submitted to initiate the "Request Access" request. This will serve as the basis for identity verification. Dataset Structure TeethAlign3D/ ├── data_738/ ├── data_739/ ├── ... ├── data_794/ │ ├── crown/ │ │ ├── up/ # Maxillary tooth crown model (from the labial perspective, from left to right) │ │ │ ├── 1.stl # The third molar on the upper left │ │ │ ├── 2.stl # The second molar on the upper left │ │ │ ├── 3.stl # The first molar on the upper left │ │ │ ├── 4.stl # The second premolar on the upper left │ │ │ ├── 5.stl # The first premolar on the upper left │ │ │ ├── 6.stl # The canine on the upper left │ │ │ ├── 7.stl # The lateral incisor on the upper left │ │ │ ├── 8.stl # The central incisor on the upper left │ │ │ ├── 9.stl # The central incisor on the upper right │ │ │ ├── 10.stl # The lateral incisor on the upper right │ │ │ ├── 11.stl # The canine on the upper right │ │ │ ├── 12.stl # The first premolar on the upper right │ │ │ ├── 13.stl # The second premolar on the upper right │ │ │ ├── 14.stl # The first molar on the upper right │ │ │ ├── 15.stl # The second molar on the upper right │ │ │ └── 16.stl # The third molar on the upper right │ │ └── down/ # Mandibular Tooth Crown Model (from the labial view, from right to left) │ │ ├── 17.stl # The third molar on the lower right │ │ ├── 18.stl # The second molar on the lower right │ │ ├── 19.stl # The first molar on the lower right │ │ ├── 20.stl # The second premolar on the lower right │ │ ├── 21.stl # The first premolar on the lower right │ │ ├── 22.stl # The canine tooth on the lower right │ │ ├── 23.stl # The lateral incisor on the lower right │ │ ├── 24.stl # The central incisor on the lower right │ │ ├── 25.stl # The central incisor on the lower left │ │ ├── 26.stl # The lateral incisor on the lower left │ │ ├── 27.stl # The canine tooth on the lower left │ │ ├── 28.stl # The first premolar on the lower left │ │ ├── 29.stl # The second premolar on the lower left │ │ ├── 30.stl # The first molar on the lower left │ │ ├── 31.stl # The second molar on the lower left │ │ └── 32.stl # The third molar on the lower left │ ├── tooth/ │ │ ├── up/ # Maxillary Closed Tooth Model │ │ └── down/ # Mandibular Closed Dental Model │ └── target_matrix.txt # Tooth Arrangement Target Transformation Matrix └── ... Data Content Description 1. Dental Crown Model (crown/) Format: STL file format Content: Triangular mesh model of the dental crown 2. Complete Tooth Model (tooth/) Format: STL file format Content: Complete closed triangular mesh model of the tooth 3. Target Matrix File (target_matrix.txt) Format: Text file Content: Target transformation matrix for each tooth after tooth arrangement Format Description: Tooth Number R00 R01 R02 0 R10 R11 R12 0 R20 R21 R22 0 Tx Ty Tz 1 Where: - Tooth Number: Corresponding to the number of the STL file (1 - 32) - M11 - M44: 16 elements of the 4×4 homogeneous transformation matrix Treatment of Missing Teeth: For missing teeth, the row with the tooth number will not be included in target_matrix.txt. Tooth Numbering System Maxillary Teeth (up/) - Numbered from left to right from the labial view - 1 - 8: Left - side teeth (from the left third molar to the left central incisor) - 9 - 16: Right - side teeth (from the right central incisor to the right third molar) Mandibular Teeth (down/) - Numbered from right to left from the labial view - 17 - 24: Right - side teeth (from the right third molar to the right central incisor) - 25 - 32: Left - side teeth (from the left central incisor to the left third molar) Dataset Scale - Number of Cases: 837 cases (data_001 to data_837) - Each Case Includes: - Dental Crown STL files (up to 16 in the maxilla: 1 - 16, up to 16 in the mandible: 17 - 32) - Closed Tooth STL files (up to 16 in the maxilla: 1 - 16, up to 16 in the mandible: 17 - 32) - 1 target matrix file (contains transformation matrices for each tooth) Note:Our dataset supports and contains missing teeth. Their STL files and matrix rows in the target matrix are empty. Application Scenarios 1. Orthodontic Treatment Planning: Optimize tooth alignment based on the target matrix 2. 3D Visualization: 3D display and analysis of tooth models 3. Biomechanical Analysis: Stress analysis of complete tooth models 4. Aesthetic Evaluation: Morphological aesthetic analysis of dental crown models 5. Machine Learning Training: Used for training AI models such as tooth segmentation and registration Technical Specifications - File Format: STL (Stereolithography) - Mesh Type: Triangular mesh - Coordinate System: Right - hand coordinate system - Unit: Millimeter (mm) - Accuracy: High - precision medical - grade scanned data Precautions for Use 1. Data Integrity: Ensure that the crown, tooth, and target_matrix.txt files for each case are complete. 2. File Association: The STL file number corresponds one - to - one with the tooth number in target_matrix.txt. 3. Coordinate System: All models use a unified coordinate system to facilitate the application of transformation matrices. 4. Missing Teeth Situation: There may be missing teeth in the case, the corresponding STL file may not exist, and the transformation matrix of the tooth will not be included in target_matrix.txt. 5. Data Verification: It is recommended to check the integrity of the STL file for each case before use to ensure the existence of the required tooth models. This dataset is only for academic research. Please abide by relevant laws, regulations, and ethical guidelines.
问问这篇 Paper
智能体会读完全文。
Lune 把这篇 Paper 索引到了每一个公式,引用它的顶会 Paper 也一样。你提问,回答直接引用原文。
它引用的顶会 Paper8
- Swin Transformer: Hierarchical Vision Transformer using Shifted WindowsZe Liu, Yutong Lin, Yue Cao, Han Hu 等ICCV 2021 · 被引用 31,683 次
- An Image is Worth 16x16 Words: Transformers for Image Recognition at ScaleAlexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn 等ICLR 2021 · 被引用 21,477 次
- Point Transformer V2: Grouped Vector Attention and Partition-based PoolingXiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu 等NeurIPS 2022 · 被引用 924 次
- OctFormer: Octree-based Transformers for 3D Point CloudsPeng-Shuai WangSIGGRAPH 2023 · 被引用 123 次
- Point TransformerHengshuang Zhao, Li Jiang, Jiaya Jia, Philip H. S. Torr 等ICCV 2021 · 被引用 23 次
相关 Paper
- TeethGenerator: A Two-Stage Framework for Paired Pre- and Post-Orthodontic 3D Dental Data GenerationChangsong Lei, Yaqian Liang, Shaofeng Wang, Jiajia Dai 等ICCV 2025 · 被引用 1 次
- TAlignDiff: Automatic Tooth Alignment assisted by Diffusion-based Transformation LearningYunbi Liu, Enqi Tang, Shiyu Li, Hui Shuai 等CVPR 2026
- Teeth-SEG: An Efficient Instance Segmentation Framework for Orthodontic Treatment Based on Multi-Scale Aggregation and Anthropic Prior KnowledgeBo Zou, Shaofeng Wang, Hao Liu, Gaoyue Sun 等CVPR 2024 · 被引用 9 次
- Collaborative Tooth Motion Diffusion Model in Digital OrthodonticsYeying Fan, Guangshun Wei, Chen Wang, Shaojie Zhuang 等AAAI 2024 · 被引用 11 次
- Segmenting Maxillofacial Structures in CBCT VolumesFederico Bolelli, Kevin Marchesini, Niels van Nistelrooij, Luca Lumetti 等CVPR 2025
