OCR module is not usable - uses huge amout of memory and is incredible slow

Hello Support,
We have a java-total license and have built a proof of concept application to test your modules word, cells (excel), slides (powerpoint), diagram (visio), pdf (conversion cgm to png) and ocr.

All of these modules are more or less working, but ocr module is not usable:

  • Uses huge amout of memory
  • Is incredible slow (for one image page about 10 seconds on my Macbook pro M1-max and more than 30 seconds within Openshift-Pod.
    What is the reason for this behaviour?
    Currently, we use software from Leadtools which works ok (about 1 second for one page).

@alexey.butalov

Can you please provide more details about the image types you are processing with the OCR module and the specific version of Aspose.OCR you are using?

We are using Version 24.10.0 of Aspose-ocr.
It’s very strange:

  • locally with Macbook M1-max some images take 2-3 seconds, others take 15 seconds or more
  • on Openshift-Cluster same images takes minutes or service (pod) crashes, due to exceeded memory (1,5 GB are allowed)

Our code:
AsposeOCR api = new AsposeOCR();

  // Set recognition options
  RecognitionSettings settings = new RecognitionSettings();
  settings.setLanguage(Language.Eng);

  settings.setDetectAreasMode(DetectAreasMode.PHOTO);

  OcrInput images  = new OcrInput(InputType.SingleImage);
  images.add(file.getInputStream());

  // Fast recognize images
  long d = System.currentTimeMillis();
  List<RecognitionResult> results = api.Recognize(images, settings);
  d = System.currentTimeMillis() - d;

Only DetectAreasMode.PHOTO delivers result.

Image is a painting of an aircraft area where some access panels are shown.

A320_test1.png (8.0 KB)

@alexey.butalov

We have opened the following new ticket(s) in our internal issue tracking system and will deliver their fixes according to the terms mentioned in Free Support Policies.

Issue ID(s): OCRJAVA-405

You can obtain Paid Support Services if you need support on a priority basis, along with the direct access to our Paid Support management team.