Build orientation for dental parts predicted by machine learning models
Predicting build orientation for SLM dental parts: a comparison of rotation representations and direct vector regression
Artificial IntelligenceComputer Vision and Pattern Recognition
Summary
Choosing the best direction to build dental parts with 3D laser printing is usually done by hand. The authors trained computer models to predict this direction automatically using images and 3D scans of teeth parts, learning from technicians' past choices. They tested different ways to represent rotations to see which worked best, finding that some approaches depend on the type of model used and using small adjustments at test time significantly improves accuracy. This helps make the orientation process less manual and more consistent.
What this means in practice
- •For dental manufacturing technicians: Automatically suggest build orientation for 3D-printed dental parts to reduce manual setup and improve consistency.
- •For 3d printing workflow engineers: Integrate machine learning models to predict part orientation from images or point clouds to optimize additive manufacturing processes.
Authors
Felix Schmalzel, Reimar Waitz, Moritz Kronberger, Thorsten Schöler
Abstract
Build orientation for selective laser melting (SLM) manufacturing of dental parts is usually chosen manually by technicians. We treat orientation prediction as supervised machine learning of the part's up-axis from technician-labeled production data, and test which rotation representations produce the best results. Using $n\approx2400$ patient-specific dental parts, we trained a ResNet-50 multi-view image backbone and a PointNeXt-S point-cloud backbone, both pretrained and fine-tuned end-to-end, on 13 up-axis representations spanning six classical $SO(3)$ parameterizations and seven representations defined directly on the unit sphere $S^2$. We report the geodesic angular error between predicted and ground-truth up-axis on a test set, with and without test-time augmentation (TTA) over $K=21$ known rotations. With TTA, the octahedral map achieves the lowest mean angular error ($10.6^\circ$, ResNet-50). The three lowest-error results overall are direct $S^2$ representations, though this may reflect label noise in the unsupervised in-plane component of the $SO(3)$ targets rather than a topological advantage. von Mises-Fisher collapses to a near-constant prediction when trained with PointNeXt-S but not with ResNet-50. TTA reduces mean angular error by 31-73 % across almost every representation and backbone. Overall, test-time augmentation over a small set of known rotations is the most consistent driver of accuracy, whereas the best-performing representation is strongly backbone-dependent.