DarSIA GUI Guide#
A from-scratch, beginner-friendly tutorial for the DarSIA GUI, for people
picking up DarSIA for the first time. Every TOML snippet is taken from
sample_config.toml (next to this guide), the running example throughout.
Warning
This guide is a work in progress. The screenshots are placeholders and will be replaced with real captures of the app.
Note
sample_config.toml is a complete, realistic FluidFlower configuration; only the
file-system paths (/path/to/experiment/...) are placeholders. To use it on your
own experiment, edit the paths in [data], [protocols], [labeling],
[facies] and [depth], and re-run Setup and Calibration from scratch (the crop
corners and calibration values are specific to one rig, camera and lighting).
Who this is for#
You know Python and are comfortable editing a text file, but you have never opened DarSIA before. You want to go from “installed DarSIA” to “produced a mass map from a FluidFlower image” without reading the source code first.
How the guide is organized#
Getting started — install, launch the GUI, tour the main window, and the five-phase mental model (Setup → Calibration → Analysis → Helper → Comparison).
Quickstart walkthrough — the core tutorial. Load
sample_config.tomland walk through every tab with screenshots and the matching config snippet.Core concepts — the ideas that make DarSIA configs click: config layering, data selectors, the ROI registry, color embeddings, and the format registry.
Configuration reference — an annotated, field-by-field reference, grounded in
sample_config.toml.Advanced analysis — segmentation styling, thresholding, finger detection, expert-knowledge ROI constraints, and the auto/manual color calibration modes.
Troubleshooting & FAQ — known issues and the questions beginners hit first.
Prerequisites#
Python 3.12+ and uv
DarSIA installed (see Getting started)
A FluidFlower image set — the walkthrough uses
sample_config.toml, but the same steps apply to your own data once you swap the paths in[data]
Start with Getting started.