← All docs

Annotations

Annotating is how your expertise enters the platform: you assign labels to particles, and those labels train and correct the models.

Labels and label domains

A particle can be labeled along several independent axes at once, each a

— taxonomy, health, morphology, or a custom domain. Each domain has its own label set, so a particle can be "Chlorella" (taxonomy) *and* "healthy" (health) simultaneously.

Certainty

Every records a

— **definite**, **likely**, or **uncertain** (the default is *likely*). It's real information, not a speed bump: downstream training and review can weigh a confident label differently from a tentative one.

Effective label

For any domain, the is resolved by precedence: your annotation wins over a model prediction, which wins over “unclassified.” That’s the answer the app shows for the particle in that domain.

Annotation vs prediction. A prediction comes from a model in a classification pipeline; an annotation is your human judgment. When both exist, the annotation takes precedence in the effective label.

Quick start

  1. Open a sample or collection and select particles.
  2. Pick a label domain and apply a label with the right certainty.
  3. Apply your changes — annotations propagate into the effective labels.