Package com.itextpdf.pdfocr.onnxtr.recognition

Class EndOfStringPostProcessor

java.lang.Object

com.itextpdf.pdfocr.onnxtr.recognition.EndOfStringPostProcessor

All Implemented Interfaces:

IRecognitionPostProcessor

       public class EndOfStringPostProcessor extends Object implements IRecognitionPostProcessor
      

Implementation of a text recognition predictor post-processor, used for OnnxTR non-CRNN model outputs.

This assumes there is an end-of-string token just after the vocabulary. You can specify additional tokens afterward, but they are not used in the processing. No same character aggregation is done. Output is read till an end-of-string token in encountered.

Constructor Summary

Constructors

Constructor	Description
EndOfStringPostProcessor()	Creates a new post-processor with the default vocabulary.
EndOfStringPostProcessor(Vocabulary vocabulary)	Creates a new post-processor without any additional tokens.
EndOfStringPostProcessor(Vocabulary vocabulary, int additionalTokens)	Creates a new post-processor.

Method Summary

Modifier and Type	Method	Description
int	labelDimension()	Returns the size of the output character label vector.
String	process(FloatBufferMdArray output)	Process ML model output and return recognized characters as string.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

EndOfStringPostProcessor

             public EndOfStringPostProcessor
             (Vocabulary vocabulary, int additionalTokens)
            

Creates a new post-processor.

Parameters:

vocabulary - vocabulary used for the model output (without special tokens)

additionalTokens - amount of additional tokens in the total vocabulary after the end-of-string token

EndOfStringPostProcessor

             public EndOfStringPostProcessor
             (Vocabulary vocabulary)
            

Creates a new post-processor without any additional tokens.

Parameters:

vocabulary - vocabulary used for the model output (without special tokens)

EndOfStringPostProcessor

             public EndOfStringPostProcessor()
            

Creates a new post-processor with the default vocabulary.

Method Details

process

             public String process
             (FloatBufferMdArray output)
            

Process ML model output and return recognized characters as string.

Specified by:

process in interface IRecognitionPostProcessor

Parameters:

output - raw output of the ML model

Returns:

recognized characters as string

labelDimension

             public int labelDimension()
            

Returns the size of the output character label vector. I.e. how many distinct tokens/characters the model recognizes.

Specified by:

labelDimension in interface IRecognitionPostProcessor

Returns:

the size of the output character label vector