Comment on page


Label for Singular Classification


Classification labels enable single-label image classification. Only one class per image can be selected. To select multiple classes per image, use Multi Classification.
Architectures such as InceptionV3 and MobileNet take selects as an input. Common use cases include explicit filters and image search.
Classification of "Rose" predicted

Define a Classification Label

  1. 1.
    Navigate to the "Label Definitions" tab of your dataset
  2. 2.
    Type label name in the "Name" field
  3. 3.
    Choose Classification from "Type" dropdown
  4. 4.
    Under Options, type the name of a class option and press "Enter"
  5. 5.
    Repeat above to create multiple class options.
Classification labels require at least 2 class options to be defined. To make this label optional and allow the selection of a none/null state, you must include a "None" class option.
6. Under Default Option, select the desired default value.
7. Click "Save" button.
After you've defined a Classification label, you can now use it to classify images in your dataset.

Data Format

A Classification schema element creates data as one of the inputed string options.
"type": "select",
"data": {
"selected": "opt1"
"children": {}

Classifying Images

Once you've defined your Classification label, you can begin annotating your data and classifying your images.

Train a Classification Model

You can train an image classification model by selecting Classification as a Model Output Type in your SmartML configuration. Check out our walk-through on building your classification model in Plainsight: