/// <summary>
/// Initializes a new instance of the <see cref="TesseractEngine"/> class.
/// </summary>
/// <param name="language1">Language 1</param>
/// <param name="language2">Language 2</param>
/// <param name="engineMode">Engine mode.</param>
/// <param name="defaultSegmentationMode">Default segmentation mode.</param>
public TesseractEngine(RecognitionLanguage language1, RecognitionLanguage language2, TesseractEngineMode engineMode, TesseractPageSegmentationMode defaultSegmentationMode)
{
RecognitionLanguage1 = language1;
RecognitionLanguage2 = language2;
EngineMode = engineMode;
DefaultSegmentationMode = defaultSegmentationMode;
}
public string TranslateLanguage(RecognitionLanguage rl)
{
return rl switch
{
RecognitionLanguage.English or RecognitionLanguage.Default => "eng",
RecognitionLanguage.French => "fra",
RecognitionLanguage.Polish => "pol",
_ => "eng",
};
}
/// <summary>
/// Gets the tesseract name of the language.
/// </summary>
/// <param name="language">Language</param>
/// <returns></returns>
private string GetLanguageName(RecognitionLanguage language)
{
switch (language)
{
case RecognitionLanguage.English:
return("eng");
case RecognitionLanguage.Russian:
return("rus");
default:
throw new ArgumentOutOfRangeException(nameof(language), language, null);
}
}
public SearchOptions(string path, string keyword, RecognitionPrecision selectedPrecisionOCR, RecognitionLanguage selectedLanguageOCR,
bool caseSensitive, bool useOCR, bool useSubfolders, bool usePDF, bool useDOCX, bool useODT)
{
Path = path;
Keyword = keyword;
SelectedPrecisionOCR = selectedPrecisionOCR;
SelectedLanguageOCR = selectedLanguageOCR;
CaseSensitive = caseSensitive;
UseOCR = useOCR;
UseSubfolders = useSubfolders;
UsePDF = usePDF;
UseDOCX = useDOCX;
UseODT = useODT;
}
public void Initialize(Enums.Language language)
{
switch (language)
{
case Enums.Language.English:
this._language = RecognitionLanguage.English;
break;
case Enums.Language.Russian:
this._language = RecognitionLanguage.Russian;
break;
default:
throw new Exceptions.InvalidLanguageException(language.ToString());
}
}
/// <summary>
/// Gets the language names string.
/// </summary>
/// <param name="language1">Language1.</param>
/// <param name="language2">Language2.</param>
/// <returns></returns>
private string GetLanguageNamesString(RecognitionLanguage language1, RecognitionLanguage language2)
{
if (language1 == RecognitionLanguage.None && language2 == RecognitionLanguage.None)
{
return("eng");
}
if (language1 == RecognitionLanguage.None)
{
return(GetLanguageName(language2));
}
if (language2 == RecognitionLanguage.None)
{
return(GetLanguageName(language1));
}
return($"{GetLanguageName(language1)}+{GetLanguageName(language2)}");
}
string ImageToText(byte[] imageBytes, RecognitionLanguage rl, RecognitionPrecision rp)
{
try
{
using var engine = new TesseractEngine(@"./Files", _optionService.TranslateLanguage(rl), EngineMode.Default);
using var img = _optionService.TranslatePrecision(rp).Item4 == System.Drawing.Imaging.ImageFormat.Tiff ?
Pix.LoadTiffFromMemory(imageBytes) : Pix.LoadFromMemory(imageBytes);
using var pager = engine.Process(img);
return(pager.GetText().ToString());
//System.Diagnostics.Debug.WriteLine("Mean confidence: {0}", pager.GetMeanConfidence());
//System.Diagnostics.Debug.WriteLine("Text {0}", text);
}
catch (Exception ee)
{
System.Diagnostics.Debug.WriteLine("Unexpected Error: " + ee.Message);
System.Diagnostics.Debug.WriteLine("Details: ");
System.Diagnostics.Debug.WriteLine(ee.ToString());
}
return("");
}
public async System.Threading.Tasks.Task <TextMessage> ProcessAsync(VoiceMessage message)
{
switch (message.Language)
{
case Core.Enums.Language.English:
this._language = RecognitionLanguage.English;
break;
case Core.Enums.Language.Russian:
this._language = RecognitionLanguage.Russian;
break;
default:
throw new Exceptions.InvalidMessageException(message.Id, "Invalid Language: " + message.Language.ToString());
}
var apiSetttings = new SpeechKitClientOptions($"{YandexCompmnentConfig.YandexSpeechApiKey}", "MashaWebApi", Guid.Empty, "server");
using (var client = new SpeechKitClient(apiSetttings))
{
MemoryStream mediaStream = new MemoryStream(message.Vioce);
var speechRecognitionOptions = new SpeechRecognitionOptions(SpeechModel.Queries, RecognitionAudioFormat.Wav, RecognitionLanguage.Russian);
try
{
var result = await client.SpeechToTextAsync(speechRecognitionOptions, mediaStream, cancellationToken).ConfigureAwait(false);
if (result.TransportStatus != TransportStatus.Ok || result.StatusCode != HttpStatusCode.OK)
{
//Handle network and request parameters error
}
if (!result.Result.Success)
{
//Unable to recognize speech
}
var utterances = result.Result.Variants;
if (utterances.Count > 0)
{
var max = utterances[0];
foreach (var item in utterances)
{
if (item.Confidence > max.Confidence)
{
max = item;
}
}
TextMessage res = new TextMessage()
{
Id = message.Id,
Language = message.Language,
Text = max.Text
};
return(res);
}
throw new Exception("invdlid answer");
}
catch (OperationCanceledException)
{
throw new Exception("invdlid answer");
}
}
}
/// <summary>
/// Initializes a new instance of the <see cref="TesseractEngine"/> class.
/// </summary>
/// <param name="language1">Language 1</param>
/// <param name="language2">Language 2</param>
/// <param name="engineMode">Engine mode.</param>
public TesseractEngine(RecognitionLanguage language1, RecognitionLanguage language2, TesseractEngineMode engineMode)
: this(language1, language2, engineMode, TesseractPageSegmentationMode.Auto)
{
}
/// <summary>
/// Initializes a new instance of the <see cref="TesseractEngine" /> class.
/// </summary>
/// <param name="language1">Language 1</param>
/// <param name="language2">Language 2</param>
public TesseractEngine(RecognitionLanguage language1, RecognitionLanguage language2)
: this(language1, language2, TesseractEngineMode.Default)
{
}
/// <summary>
/// Create new speech recognition options.
/// </summary>
/// <param name="speechModel">The language model to use for recognition.</param>
/// <param name="audioFormat">The audio format.</param>
/// <param name="language">The language for speech recognition.</param>
public SpeechRecognitionOptions(SpeechModel speechModel, RecognitionAudioFormat audioFormat, RecognitionLanguage language)
{
SpeechModel = speechModel;
AudioFormat = audioFormat;
Language = language;
}
public Task ProcessImage(byte[] data, RecognitionLanguage language)
{
var settings = new ProcessingSettings
{
Language = language,
OutputFormat = OutputFormat.txt,
TextTypes = TextType.Normal
};
// set request
Task task = _restServiceClient.ProcessImage(data, settings);
return task;
}
