Freeform Text

Freeform text annotations give you full control over the annotation format for each image. Instead of predefined structures like bounding boxes or question-answer pairs, you write any text you want in a free-form editor. This is useful for custom JSON schemas, detailed descriptions, multi-attribute labels, or any format your training pipeline requires. This guide walks through creating freeform text annotations in Datature Vi.

Open the annotator

Go to your dataset, then click the Annotate tab. Click any image thumbnail in the bottom strip to load it onto the canvas.

1

Open the Annotator tab

From the Dataset Overview page, click the Annotator tab to open the labeling interface.

Next →

You should see

Your annotations are ready when you see annotation count matching the image count in the Dataset Overview.

Keyboard shortcuts

Annotation guidelines

These guidelines produce annotations that train well. Consistency across your dataset matters more than the format you choose.

JSON example:

{
  "description": "A solar panel array on a residential rooftop",
  "condition": "good",
  "panel_count": 12,
  "issues": ["minor dust accumulation on panels 3 and 7"],
  "orientation": "south-facing"
}

Plain text example:

Description: A solar panel array on a residential rooftop
Condition: Good
Panel count: 12
Issues: Minor dust accumulation on panels 3 and 7
Orientation: South-facing

Edit or delete annotations

To edit an annotation: Click in the Freeform panel text area and make your changes. Changes save automatically.

To clear an annotation: Select all text in the editor and delete it. The annotation is removed when the text is empty.

Chain-of-thought reasoning

<datature_think>Starting from the top shelf, I count 3 cardboard boxes.
The middle shelf has 2 plastic bins and 1 cardboard box.
The bottom shelf holds 4 plastic bins.
Total: 4 cardboard boxes, 6 plastic bins.</datature_think>

The shelf contains 4 cardboard boxes and 6 plastic bins across three levels.

Next steps