Low recognition quality - Aspose.OCR 17.6.0.0


#1

Dear Supporters,

Aspose.OCR v17.6.0.0 (.net C #)
Portuguese language resource file for Aspose.OCR for .NET 3.2.0

We are testing Aspose.OCR 17.6.0.0 with the intention of buying the product.
Unfortunately the product has demonstrated a poor quality in recognition.
The images used have 300 DPIs.

Could you please inform us if this is really a deficiency of the product?
Are we doing something wrong?
Does the product also support the Portuguese language of Brazil?
Is it possible to train this product to increase its assertiveness percentage?

Attached example of the images and results we obtained:

  • 1 - image0000001A.jpg
  • 1 - result-image0000001A.txt
  • 2 - Daniel-004-TituloEleitor.jpg
  • 2 - result-Daniel-004-TituloEleitor.txt
  • Aspose.OCR-RecognitionQuality.zip

Image 1
1 - image0000001A.jpg (398.9 KB)
Result image 1
1 - result-image0000001A.JPG (228.3 KB)

Image 2
2 - Daniel-004-TituloEleitor.jpg (671.3 KB)
Result image 2
2 - result-Daniel-004-TituloEleitor.JPG (20.9 KB)

Zip files
Aspose.OCR-RecognitionQuality.zip (1011.5 KB)

Here is also the example code used by us:

if (FileUploadInput.PostedFile != null)
{
string appPath = HttpRuntime.AppDomainAppPath;

            string fullpath = Path.Combine(appPath, "files", string.Concat(DateTime.Now.ToString("yyyyMMddHHmmss"), Path.GetExtension(FileUploadInput.PostedFile.FileName)));

            string directory = Path.GetDirectoryName(fullpath);
            string dataPath = Path.Combine(appPath, "Portuguese_language_resource_file_for_Aspose.OCR_for_.NET_3.2.0.zip");

            LoadLicense(appPath);

            if (!Directory.Exists(directory))
                Directory.CreateDirectory(directory);

            FileUploadInput.PostedFile.SaveAs(fullpath);

            // Usando region
            OcrEngine engine = new OcrEngine();

            engine.LanguageContainer.Clear();
            engine.LanguageContainer.AddLanguage(LanguageFactory.Load(dataPath));
            engine.Config.DoSpellingCorrection = true;

            engine.Image = ImageStream.FromFile(fullpath);

            if (engine.Process())
            {
                TextBox.Text = engine.Text.ToString();
            }
            else
            {
                TextBox.Text = "Error: Did not process";
            }
        }

Best regards,
Tiago.


#2

Can someone help me please?
We are urgently on the subject, my team is already realizing the purchase of Aspose.Total and we have already developed an entire project on top of that Aspose.OCR DLL.


#3

@tiago.silva

We apologize for the inconvenience faced.

We have tested the scenario in our environment while using your images and noticed that API was unable to extract text correctly. We have logged following issues in our issue tracking system for the sake of correction.

  • OCR-727
  • OCR-728

We will definitely look into details of these issues and keep you posted with the status of their correction. Please spare us little time.

Furthermore, please note that Aspose.OCR for .NET has not been updated since its 17.6 version. We are currently in process to launch latest version of the API with new features and improvements. On the other hand, Aspose.OCR Cloud is being updated on monthly basis and eligible to produce much better results. You may please try our Cloud API and in case you face any issue, please feel free to let us know.


#4

Could you please inform me your deadline for updating the Aspose.OCR for .NET DLL?

We have already developed a whole project based on these Aspose.OCR DLLs, since they are still available for use.


#5

@tiago.silva

We are currently working over implementing new features and building API components. We are afraid that we cannot share any ETA at the moment. However, we will get back to you soon with some updates. Please spare us little time.


#6

Any news on this subject?


#7

@tiago.silva

Thanks for your patience.

We would like to share with you that we are very close to release new version of the API i.e. Aspose.OCR for .NET. We are implementing a new OCR Engine in terms of the state-of-the-art AI technologies, and it is taking a lot of time to make it work properly. However, the intended release of the API may be expected within two or three months. In case of any delay or release time gets shorter than intended, we will let you know. Please spare us some time.