Package com.itextpdf.pdfocr.onnxtr.detection

Interface IDetectionPostProcessor

All Known Implementing Classes:
OnnxDetectionPostProcessor
public interface IDetectionPostProcessor
Interface for post-processors, which convert normalized, but still raw output of an ML model and returns rotated boxes with the detected objects.

Output point arrays should represent a rectangle and contain 4 points. Order of the points should be the following: Bottom-Left, Top-Left, Top-Right, Bottom-Right. This is with "bottom" and "top" referring to how it would look on an image, not Y coordinate.

Its box points are return in a different order, it will cause issues with the following steps (like orientation prediction and text recognition).

  • Method Summary

    Modifier and Type
    Method
    Description
    List
    process(BufferedImage input, FloatBufferMdArray output)
    Process ML model output for a specified image and return a list of detected objects.

  • Method Details

    • process

      List process (BufferedImage input, FloatBufferMdArray output)
      Process ML model output for a specified image and return a list of detected objects.
      Parameters:
      input - input image, which was used to produce the inputs to the ML model
      output - normalized output of the ML model
      Returns:
      a list of detected objects. See interface documentation for more information