Skip to main content

Dataset Labels

Datasets use labels whose status is Finalized and whose class is selected in the dataset settings. Draft and other non-finalized labels are excluded.

Finalize the intended label sets in Spectral Explorer before assigning their images. On the Images tab, expand an assigned image to review the label sets used in that bucket. If Clarity reports that no finalized labels were found for a selected class, finalize the labels or change the dataset classes.

Merge source classes​

In dataset settings, select Classes as usual. To merge classes, open that dropdown, select the configure button beside Search, then Merge classes…. Choose an Output class for each source. Choosing the same existing managed output class for multiple sources merges their samples in this dataset version. For example, map two plastic subclasses to one managed plastic class while keeping a third class separate: three source classes produce two model classes.

The original annotations and label-set selections stay unchanged. Reference spectra use the same mapping, and sampling limits and statistics apply to the output classes. Configure models with these output classes. Change the mapping on a draft version; finalized versions retain their mapping.