divisualisation

A napari plugin to visualise cell-tracking errors, computed via traccuracy, by lifting 2D/3D+time tracks into an interactive 3D "spacetime" view.

🆕 divisualisation is now a fully fledged napari plugin, with a stateful spacetime lifted view that integrates with regular napari workflows.

We originally introduced these visualisations to compare our results in Trackastra: Transformer-based cell tracking for live-cell microscopy to other cell tracking algorithms.

Installation

1. Please install napari as outlined here.

2. After that, install divisualisation, either:

   • from within napari via Plugins → Install/Uninstall Plugins… (search for "divisualisation"),
   • or from PyPI:

     pip install divisualisation
     

   • or the latest development version from GitHub:

     pip install git+https://github.com/bentaculum/divisualisation.git
     

Note: requires Python ≥ 3.11 and napari ≥ 0.8.

Usage

Open Plugins → divisualisation → Lift tracks & Divisualisation. The widget has two independent workflows, each in its own box:

• Lift all tracks layers — fold time into a z axis so every tracks layer rises out of the image plane into a 3D "spacetime" cone. Scrub the time slider to sweep through the cone; toggle off to restore the flat view exactly.
• Divisualisation — assign ground-truth / predicted / FN-edge / FP-edge tracks layers via the role dropdowns (auto-guessed from layer names), Compute edge errors from the GT/predicted tracks plus their labels, and lift with the error colouring. Color division edges draws each layer's parent→daughter edges as coloured tails (napari otherwise draws them in uncolourable white).

Examples

Run in ipython — each loads data into a viewer, adds the tracks and edge-error overlays, and docks the widget:

• example_2d.py — bacteria (2D+t).
• example_3d.py — C. elegans nuclei (3D+t, z scaled ×10).
• example_programmatic_2d.py — fully scripted render (no GUI): build layers, lift with SpacetimeLift, overlay errors with add_edge_error_tracks, capture a napari_animation keyframe video.