# Image Annotation

Source: https://help.zira.us/docs/ai-vision/image-annotation
Summary: Open one dataset image to add and update labels in one place.
Updated: 2026-11-03

![Image annotation page](/docs/data-sources/image-annotation-01.png)

Open a dataset, then click an image to open this page.

This page is where you review one image, work on labels, and move between images without going back.

## Auto annotation

The buttons at the top run auto annotation. Each one is a separate tool, and the small **⋮** next to
it opens that tool's own settings.

- **Build pallet** boxes the pallet, boards, stringers and nails. With no boxes yet, it finds the
  boards first.
- **Custom model** runs a model trained for this data source.
- **Generic model** finds your labels with a general vision model. No training needed.
- **Describe** lets you type what to box in plain words, like "missing boards".
- **Click to annotate** turns on click mode: click an object to box it. Press **Esc** to stop.
- **Board defects** highlights cracks and breaks on the board boxes. It is a preview and adds no boxes.
- **Paste JSON** adds boxes from pasted coordinates, or lets you edit this image's boxes as JSON.

The tool marked **A** is the last one you ran. Press **A** to run it again.

Use **Many images…** to run a tool on the next images in the dataset. It reads each image's saved
boxes first: with **Only images with no boxes yet** ticked (the default), images that already have
boxes are left alone, even ones you have not opened. An image whose boxes can't be read is skipped.
Choose how many images to work on at once (1 to 6, 3 by default).

### Only the tools you need

The page shows the tools that fit what you are annotating:

- **Vision** cameras get the model, Describe, click and paste tools.
- **Pallet images** (whole pallets) also get **Build pallet** and **Board defects**.
- **Nail crops** (one board-and-stringer crop per image) get the model, Describe, click and paste
  tools, and Custom model starts with your newest nails model.
- **Band saw boards** (band saw cameras) get **Build pallet** with the model, Describe, click and paste
  tools. Here Build pallet finds the boards and stringers and adds only the ones not boxed yet: your
  boxes, broken boards included, stay as they are, and it adds no pallet box or nails. Box a few boards
  first and it completes the rest.

The page picks this for you. To change it, click **Tools** at the top right.

## Annotate settings

Click the **⋮** next to a tool, or **Annotate settings** on the left, to change how a tool works.
The settings list the same tools as the top bar, in the same order, and each one has **Run on this
image**.

![Generic model settings](/docs/data-sources/image-annotation-02.png)

### Generic model

Use **Generic model** when you want the system to find common objects from your label list.

- choose the detection model
- choose which labels to include
- add simple label descriptions if needed
- adjust how strict the results should be

This is usually the easiest place to start.

![Custom model settings](/docs/data-sources/image-annotation-03.png)

### Custom model

Use **Custom model** when you already trained a model for this data source.

- pick the trained model
- adjust confidence
- adjust overlap filter
- set how many boxes can be added for each label

This is usually better when the generic option is not accurate enough.

![Click To Annotate settings](/docs/data-sources/image-annotation-04.png)

### Click to annotate

Use **Click to annotate** when you want help on one object at a time.

- choose the label on the left
- turn on **Click to annotate** at the top, or hold **Option**
- click the object in the image

If **Classify clicked objects** is on, the system tries to decide the label for you. If it is off, it uses the label you selected on the left.

### Zones

Zones limit **Generic model** and **Custom model** to part of the image. An include zone searches
only inside it; an exclude zone skips that area. Draw them from **Auto-annotate zones** on the left,
or from the tool's settings.

### Included labels

Use **Included labels** to choose which labels the system should work with.

If you do not need every label, keep only the ones you want. This usually gives cleaner results.

### Label descriptions

You can add short words or phrases for each label.

This helps the system understand what each label means.

### Results

You can make the result more loose or more strict.

- **Confidence**: lower finds more objects, higher keeps only stronger matches
- **Overlap filter**: removes duplicate boxes
- **Max boxes per label**: limits how many boxes are added

### When boxes are added

- **Wait for the boxes on this image** shows the result here before you move on
- **In the background** lets you move to the next image; the boxes save to the image the run started on

## View

Use the **View** section on the left to change how you look at the image.

- **Turn selection off/on** changes whether you are selecting labels
- **Show active only** shows only the label you are working on

## Right-click a box

Right-click a box (two-finger click on a Mac trackpad) to:

- **Hide** it, so you can see what is under it. It is still saved. Click **Show** at the top right
  to bring hidden boxes back.
- **Delete** it
- **Change label** to any label in the list

If the box is part of a selection, the menu changes every selected box.

## Actions

Use the **Actions** section on the left for common image work.

- **Annotate settings** opens the settings for every tool
- **Shortcuts** shows keyboard shortcuts
- **Copy labels** copies labels to other images
- **Bulk edit labels** updates labels in bulk
- **Delete image** removes the image

## Labels

Use the **Labels** section on the left to work with your label list.

- click a label to make it active
- the number next to each label shows how many of that label are in the image
- **Configure labels** opens the dataset label list

## Image list

The images on the right help you move through the dataset.

- click another image to open it
- the highlighted image is the one you are working on now
- the count at the top shows how many images are in the list

## Read next

- [Dataset Page](/docs/ai-vision/dataset-page)
- [Datasets](/docs/ai-vision/datasets)
