Skip to main content

Create a Dataset

Start a draft​

  1. Open Datasets and select New dataset.
  2. Enter the dataset name. The dataset details drawer opens on Overview.
  3. Assign a project if needed.

Assign images​

  1. Open Images.
  2. Choose Manual or Automatic under Data assignment method:
    • Manual provides separate Training, Validation, and Test groups.
    • Automatic provides Training & Validation plus Test and enables a training/validation split slider.
  3. In Available data, expand folders and select images or folders.
  4. Select Add to bucket and choose the destination group. You can also drag compatible entries between Available data and the assignment groups.

Clarity can auto-select classes, processing level, or wavelengths when the assigned training data has one unambiguous common choice. Review every auto-selected value before finalizing.

Configure the draft​

On Overview, set:

  • Wavelengths
  • Data level processed
  • at least one Class
  • optional label attributes and mask handling
  • Sampling method and Class split type
  • the number of latent variables when PCA is enabled

Use Reference spectra to add signatures when they are part of the training or validation design.

Resolve validation messages on Overview and invalid entries on Images, then follow Finalize Dataset.