Image classification

Updated 1 week ago by Alex Cota

Overview

The image classification tool only produces a semantic classification for the image as a whole and does not include any geometric annotation or pixel classification information. There are four ways for labelers to provide a value for image classification: radio, checklist, dropdown, and free-form text classification.

Radio

Use the radio format if you would like a labeler to choose a single answer from a selection of answer choices.

The nested radio format allows the labeler to select multiple answers from a nested taxonomy.

Configuration

//Radio
"classifications: [
{
"featureId": "",
"schemaId": "",
"title": "",
"value": "",
"answer": {
"featureId": "",
"schemaId": "",
"title": "",
"value": ""
}
}
]

Checklist or Dropdown

Use the checklist or dropdown format if you would like a labeler to choose multiple answers from a selection of answer choices.

The nested dropdown allows labelers to select answers from a nested taxonomy.

Configuration

//Checklist OR Dropdown
"classifications: [
{
"featureId": "",
"schemaId": "",
"title": "",
"value": "",
"answers": [
{
"featureId": "",
"schemaId": "",
"title": "",
"value": ""
}
]
}
]

Text classification

Use the text classification format if you would like a labeler to enter a free-form text answer in response to a classification question.

There are various cases where you might want to use free-form text input such as for optical character recognition (OCR) tasks. Often OCR tasks have use the bounding box tool along with the nested free-form text classification tool.

Configuration

// Text classification
"classifications": [
{
"featureId": "",
"schemaId": "",
"title": "",
"value": "",
"answer": ""
}
]

Was this section helpful? Leave your feedback below.


How did we do?