First run¶

This page takes you from a freshly installed IQUANA to one measured object.

1. Start the services¶

./iquana.sh start

./iquana.sh status shows what is running, on which port, and at which commit. If something is missing, its log is in logs/ — see Command line.

Open http://localhost:3000.

2. Create the first account¶

The first account can always be created, whether or not self-registration is allowed — a fresh installation is never locked out of itself. Create it now; it becomes the platform administrator.

If other people will use this installation, read Configuration → identifying your instance before handing out the address, and decide deliberately whether strangers may register themselves. Self-registration is off by default.

3. Create a dataset and upload images¶

A dataset is the unit everything else hangs off: images, labels, annotations, quantification profiles and membership all belong to one dataset.

See Create a dataset.

Recording slot — dataset-upload
Create a dataset, upload images, add typed per-image metadata, filter the gallery by it.
10–14 s · dark theme · 1280×800

4. Define at least one label¶

You cannot save an annotation without a label to attach it to. Labels nest into parent–child trees, so start with something coarse — you can refine it later without discarding the annotations above it.

See Build a label hierarchy.

5. Annotate one object¶

Open an image in the annotation workspace, pick the Point tool (P), click on an object, and press Enter to run the model on the prompt. Accept the result and assign a label with L.

The full prompt vocabulary — positive and negative points, boxes, polygons and freehand — is in Annotate with AI. Press ? at any time for the keyboard shortcuts.

6. Calibrate, so the numbers mean something¶

Until you calibrate, every measurement is in pixels. Calibration maps pixels to a physical length unit, and can be set per dataset when your acquisition setup is consistent.

See Calibrate to physical units.

7. Quantify and export¶

With at least one labelled, calibrated object, the quantification page will compute metrics for it and let you export the result.

See Quantify and export and the metric reference.

Where to go next¶

You have been through the loop once. From here:

Scale it up — Run batch inference across the whole dataset instead of image by image
Bring in other people — Manage access
Improve the model — Train a model on your corrected annotations