Remove TODOs to DEVSIX-9153

Author: AnhelinaM

pdfocr-api/src/main/java/com/itextpdf/pdfocr/TextInfo.java

@@ -40,11 +40,6 @@ public class TextInfo {
 
     /**
      * {@link Rectangle} describing text bbox (lower-left based) expressed in points.
-     *
-     * <p>
-     * TODO DEVSIX-9153: mark this on breaking changes page. Why not return rectangles in image pixels?
-     *  Seems odd, that an OCR engine should be concerned by PDF specific. It would make sense for an engine
-     *  to return results, which could be directly applied to images inputs instead.
      */
     private Rectangle bboxRect;
 

pdfocr-onnxtr/src/main/java/com/itextpdf/pdfocr/onnxtr/OnnxTrProcessor.java

@@ -93,15 +93,15 @@ Map<Integer, List<TextInfo>> doOcr(List<BufferedImage> images, OcrProcessContext
                     eventHelper.getSequenceId(), null, eventHelper.getConfirmationType());
             eventHelper.onEvent(event);
             /*
-             * TODO DEVSIX-9153: Potential performance improvement (at least for GPU).
+             * Potential performance improvement (at least for GPU).
              *
              * There is a potential for performance improvements here. Currently, this mirrors the
              * behavior in OnnxTR/DocTR, where inputs for orientation and recognition models are
              * aggregated per input image.
              *
              * But, most of the time, this will not be enough to saturate the batch size fully.
              * Ideally, we should process all text boxes together, regardless of the origin image,
-             * and then separate the results afterwards.
+             * and then separate the results afterward.
              */
             BufferedImage image = images.get(imageIndex);
             List<Point[]> textBoxes = textBoxGenerator.next();

pdfocr-onnxtr/src/main/java/com/itextpdf/pdfocr/onnxtr/detection/OnnxDetectionPostProcessor.java

@@ -92,10 +92,9 @@ public List<Point[]> process(BufferedImage input, FloatBufferMdArray output) {
         final int height = output.getDimension(1);
         final int width = output.getDimension(2);
         final List<Point[]> boxes = new ArrayList<>();
-        // TODO DEVSIX-9153: Ideally we would want to either cache the score mask (as model
-        //       dimensions won't change) or use a smaller mask with only the
-        //       contour. Though based on profiling, it doesn't look like it is
-        //       that bad, when it is only once per input image.
+        // Ideally we would want to either cache the score mask (as model dimensions won't change)
+        // or use a smaller mask with only the contour. Though based on profiling, it doesn't look
+        // like it is that bad, when it is only once per input image.
         try (final Mat scoreMask = new Mat(height, width, CvType.CV_8U, new Scalar(0));
              final MatVector contours = findTextContours(output, binarizationThreshold)) {
             final long contourCount = contours.size();