Data Notice
Data Source Notice
ToneFinder may reference public datasets during parts of model development and evaluation. This page briefly explains the source datasets, rights holders, and how those references are used in the service.
Last updated August 24, 2026
1. Source Datasets
Some ToneFinder model development and evaluation workflows may reference public datasets provided through AI Hub. In the current service, these references are especially relevant to skin lesion classification model training and performance checks.
2. Rights Notice
Rights to source datasets referenced through AI Hub belong to the National Information Society Agency (NIA) and the relevant rights holders or participating institutions. ToneFinder does not claim ownership over the raw datasets, and source data usage remains subject to the original provider policies.
3. How We Use Them
The service delivers model outputs and result screens rather than raw dataset files. ToneFinder does not directly provide or sell source datasets, and any separate use of the underlying data should follow AI Hub or the relevant rights holder guidance.
4. Notes
Skin lesion classification and personal color outputs in the service are AI-generated predictions. Data providers do not guarantee individual service results, so please refer to the original dataset policies when needed.
5. User Analysis Data
Photos submitted by users for analysis are handled separately from the public training datasets described above. The photo quality pre-check runs in the browser and is not uploaded separately, but starting the actual analysis sends the selected photo to the server to generate a result. Product photos used by the shopping color matcher stay on the device and are processed locally with Canvas.
Original personal-color photos and generated overlays are stored in the configured server storage to provide device-specific result history. The application currently has no automatic image-deletion period, and expiration of the 30-day device session does not delete stored data. Permanent deletion removes the stored R2 images idempotently first, then hard-deletes the database record. If R2 deletion fails, the history entry and share remain so the same request can be retried; the request can also be retried if database cleanup fails after image removal. The default seven-day public share contains no photo or overlay.