Comment on page
Multi Classification
Label for Multi-Class Image Classification
A Multi Classification schema element creates data as any number of the inputed string options. If no selection is made, or a default is not set, the "selected" value will be an empty array.
Multi Classification labels enable multi-label image classification, whereas Classification enables single-label image classification.
Architectures such as InceptionV3 and MobileNet can be converted to support multiple selections. Common use cases include attribute identification.
Classifications of "Van Gogh, Painting, Fine Art, Post-Impressionism" predicted
Plainsight does not support model training for this label type. This label type can be exported in Plainsight format or CreateML Classifier for training outside of Plainsight.
- 1.Navigate to the "Label Definitions" tab of your dataset
- 2.Type label name in the "Name" field
- 3.Choose Multi Classification from "Type" dropdown
- 4.Type the name of the class and press "Enter"
- 5.Repeat above to create multiple classes
- 6.Optional: Choose default class labels
- 7.Click "Save" button
Once you have defined a Multi Classification label type, you can use it to classify your images into one or more classes:
A Multi Classification schema element creates data as any number of the inputed string options. If no selection is made, or a default is not set, the "selected" value will be an empty array.
{
"type": "multiSelect",
"data": {
"selected": ["red", "green"]
},
"children": {}
}
Once you've defined your Multi Classification label, you can begin annotating your data and classifying your images.
Plainsight doesn't currently support SmartML model training with multi classification labels. Singular classification with the Classification label type, however, is supported with SmartML training. Both of these label types can be exported using the CreateML Classifier format and trained outside of Plainsight.
Last modified 1yr ago