DentalOPEN SOURCE

Visual contribution guide

From images to reviewable models

In preparation: file upload and AI model generation are not available yet.

Do not send clinical images through GitHub, public links or the current text form.

Private photo workspace · Özel fotoğraf alanı ↗

1 · Tooth photographs

Keep each tooth in its own image set. Record the FDI number if known, view directions, and missing or damaged regions. Do not guess unknown information.

Include front, back, side, chewing-surface and root-tip views with overlapping intermediate angles. Keep focus and lighting consistent. Include a calibration reference if available; otherwise scale remains unknown.

Photographs support visible external surfaces; they do not measure internal pulp or canals. Do not present AI-enhanced images as original evidence.

2 · Radiographs and volumes

A 2D radiograph and CBCT are different inputs. A proposed 3D shape from one radiograph includes inference and is not presented as measured anatomy.

Prepare only existing material you are authorized to contribute. This project does not request new imaging or tooth extraction.

3 · Private review and permissions

Files will first undergo security, rights and human privacy review in a private area. Image text and metadata will be checked separately.

Processing, AI inference, publication and model-training permissions will be separate. Contributing does not automatically authorize publication or AI training.

4 · Candidate model and expert review

AI output will be presented as a candidate with its sources, method, version, limitations and inferred regions. Experts will record a reasoned acceptance, revision request or rejection.

Accepted models will remain open to criticism. New findings will reference the relevant version; changes will create a new version and review. AI cannot approve its own model.

Until the controlled upload channel opens, this page provides guidance only.

Full capture and contribution guide

Generation and expert-review design

Candidate classes and publication limits

Candidates are classified by their evidence basis. No class alone constitutes scientific acceptance.

Candidate classes and publication limits
OBSERVATION_DERIVED
Observation-derived candidate with calibration not yet verified.
RECONSTRUCTED_FROM_CALIBRATED_IMAGING
Candidate derived from volumetric imaging with a referenced calibration record.
MULTIVIEW_SURFACE_RECONSTRUCTION
External surface reconstruction supported by multiple photographs.
AI_INFERRED_HYPOTHESIS
Hypothesis inferred using learned priors or a generative model.
HYBRID
Candidate combining image-supported and inferred regions.

Models based only on 2D radiographs cannot be published as that specimen’s 3D anatomy or as canonical models; they may only be considered as illustrative research hypotheses or model-prior visualizations.

Each region will show its source view, registration record, support type, confidence-score meaning when available, and expert-review status.