Skip to main content
For each classifier, create a training batch, add training images, and assign reference classes.

​
Create a training batch

1

Open the classification training batch view

Do one of the following:
  • Click the Open Classifier Training Batches button on the toolbar.
  • Select Open Classifier Training Batches in the View or Classification Training menu.
  • Press Ctrl+Alt+Shift+B.
2

Create a batch

Right-click the list of batches and click New Batch on the shortcut menu.
3

Set the classification profile

Select Change Classification Profile in the Classification Training menu or the batch shortcut menu. For more information, see Classification profiles.
4

Set the precision/recall balance

Select Change Recall/Precision Priority in the Classification Training menu or the batch shortcut menu. For more information, see Precision/recall balance.
5

Add training images

Do one of the following:
  • Click Load Images from Folders on the toolbar or select it in the File menu, and then choose the folder that contains the subfolders of images. Each subfolder should contain images of a single class only.
  • Click Load Images in the batch shortcut menu or the File menu.
6

Assign a reference class

Use one of the following methods:
  • Click Set Class in the Classification Training menu, or right-click a document and click Set Class. Select a reference class, or create one and link it to a section or section variant. You can use sections or section variants as reference classes. The reference class must have the same name as the corresponding Document Definition.
  • If you added the images with Load Images from Folders, assign classes by subfolder name: select the images and click the Set Class Based on Subfolder Name button on the toolbar, or select Classification Training → Set Class Based on Subfolder Name.
Image pre-processing applies only to pages added with Load Images. Pages added with Load Images from Folders are processed according to the Project properties. For the Image or Combined profile, use training images with the same color scheme (for example, black-and-white) as the images you will classify.
If a class name matches a Document Definition section name, they link automatically when you click Autocorrect in the Class Mapping dialog box. For more information, see Map classes to Document Definition sections.

​
Assign reference classes from classification results

If you have many unsorted images for training, train the classifier on some of them and assign reference classes to the rest based on the results.
1

Set some reference classes manually

Assign reference classes to some of the images with Set Class.
2

Train the classifier

Click the Train button on the toolbar, or select Train in the Classification Training menu or the shortcut menu.
3

Classify the remaining images

Select the images that have no reference class and click the Classify button on the toolbar, or select Classify in the Classification Training menu or the shortcut menu. Each image is assigned a result class based on the trained classifier.
4

Promote result classes to reference classes

With the images selected, click Set Class Based on Classification Result in the Classification Training menu or the shortcut menu. Each image is assigned its result class as its reference class. Fix any image that received an incorrect result class.
To reuse existing recognition results when matching Document Definitions, select Prefer settings from batch type on the Classification tab of the training batch properties (batch shortcut menu → Properties). This synchronizes full-text recognition settings so documents are not recognized again, which saves classification time.