Quick start

This walk-through takes you from a folder of images to an exported dataset in about five minutes.

1. Create a project

Launch lannotex. On the Home screen, pick the pipeline that matches your task:

Pipeline	Use it for	Tools shown
Classification	Tagging whole images (cat, blurry, night)	Class checkboxes only
Object detection	Locating objects	Bounding boxes
Segmentation	Tracing exact outlines	Polygons
Combined pipeline	All of the above	Everything

Name the project, choose a folder for it, and pick Solo (offline, the default) or Team. Lannotex creates one file there: project.sqlite3.

You can change the enabled tasks later from Project → Project tasks….

2. Import images

Use + Folder or + Files in the Dataset panel, or press Ctrl+I. Lannotex scans folders recursively for JPEG, PNG, and WebP files and checks each one. Damaged files are reported, not imported.

Images are referenced, not copied. The project stores each path relative to the project folder, so keep the project folder and image folder in the same place relative to each other, or use Project → Relink current image… after moving a file.

3. Add labels

Type a label name in Labels & annotations and press +. Rename or remove labels at any time. Renaming updates every existing annotation.

4. Annotate

• Classification: tick the classes that apply. Tick No target objects in this image for true negatives.
• Boxes: press B, select a label, and drag on the image.
• Polygons: press P, click to place vertices, and press Enter to finish.

Press V to select and edit shapes. The full list of controls is in Annotating.

5. Save and move on

Press Ctrl+Enter to Save & next. Your work is also autosaved to a recovery draft as you go, so closing the app never loses edits.

6. Export

Choose Export (top right), pick a format and a destination folder. Lannotex writes a new Lannotex-export-xxxxxxxx folder, so it never overwrites a previous export. Images with problems are skipped and listed. See Import & export.

Check before you export

View → Analysis shows label counts and lists any image with a validation problem, a missing source file, or an unsaved draft.