Recognition

Recognition requires an input image, a page segmentation for that image (both can be supplied as through a single XML files), and a text recognition model file. In particular there is no requirement to use the page segmentation algorithm contained in the segment subcommand or the binarization provided by kraken.

It is possible to apply different recognition models to different parts of a segmentation, for example if the lines in the segmentation are annotated by language, script, or hand:

$ kraken -i ... ... ocr -m Grek:porson.mlmodel -m Latn:antiqua.mlmodel

All lines having the line type Grek (polytonic Greek) will be recognized using the porson.mlmodel model while lines with Latn type will be fed into the antiqua.mlmodel model. Obviously this requires the segmentation model to output lines with these line types. It is possible to define a fallback model that other text will be fed to:

$ kraken -i ... ... ocr -m ... -m ... -m default:porson.mlmodel

It is also possible to disable recognition on a particular script by mapping to the special model keyword ignore. Ignored lines will still be serialized but will not contain any recognition results.