Layout classification

Layout classification relies on a single project-wide Layout Classifier that looks at the physical layout of a document to determine the classification result. Use layout classification when a document has a consistent layout, such as static forms. If the first page of a document is easily recognized, it is possible to classify other documents with that same layout. However, this type of classification relies exclusively on the training documents in your Classification Set. Without training documents, the Layout Classifier cannot recognize any document.

The Classification Set requires permanently stored black and white images for layout classification. After adding a color or grayscale image to the Classification Set, and after training, a bi-tonal version of the image is created and stored in a subfolder of the project directory.

Each class in your project requires a few training documents. For the best results, add a single training document for a class using layout classification. Next, train your project and then test your classification results. If you get high confidence results with a single training document, there is little need for additional training documents. If the results are unsatisfactory, add training documents one-by-one until the classification results are satisfactory.

The layout classification result is accepted as the final classification result in the classification order if there is one. If the Layout Classifier cannot determine a classification result, content classification is performed.