Skip to main content

Annotations

Biom provides a full annotation system for marking up your data with shapes, text, segmentation masks, and comments.

Annotation types

TypeDescription
PointSingle point marker
LineLine segment between two points
RectangleRectangular region
CircleCircular region
PolygonMulti-point closed polygon
FreehandFreehand drawn shapes
TextText labels at specific positions
MaskSegmentation masks from model outputs
All annotations are stored in world coordinates, ensuring they stay aligned with the image regardless of zoom or pan.

Annotation scoping

Annotations can be scoped to:
  • Per Z-slice — visible only on a specific Z-plane
  • Per time-point — visible only at a specific time
  • All slices — visible across the entire Z-stack

Annotation features

Styling

  • Fill color, stroke color, stroke width, and opacity
  • Per-annotation style customization

Organization

  • Z-order management — bring to front, send to back, move up/down
  • Per-pane visibility — the same annotation can be shown or hidden independently per pane
  • Copy across documents — copy annotations between different files

Segmentation metadata

Annotations from model outputs (like SAM3) include:
  • Instance ID
  • Class label
  • Confidence score
  • Model name that generated the annotation

Document model

Annotations are organized into documents (collections):
  • Documents can be scoped to a tab or pane
  • Multiple documents per pane, switchable
  • Persisted locally in IndexedDB

Comments

Biom supports threaded comments attached to specific locations in your images:
  • Position-linked — comments are tied to a specific view state and location
  • Threaded replies — reply to comments to create discussions
  • Resolve/unresolve — mark comments as resolved when addressed
  • Per-pane modals — comment panels are independent per pane
  • Server persistence — comments sync to the server for shared workspaces
  • Read-only in shares — viewers can read but not add comments on shared links

Keypoint labeling

Biom includes a dedicated keypoint annotation workflow designed for creating DeepLabCut training datasets:
  • Session-based — create labeling sessions with defined keypoint schemas
  • Frame navigation — step through video frames or image stacks
  • Individual keypoint selection — click to place each keypoint
  • Undo/redo — full history stack for corrections
  • Color palettes — customizable keypoint colors
  • Export — export labeled keypoints for model training