public OcrRequestParams(ScannedImage.Snapshot snapshot, IOcrEngine ocrEngine, OcrParams ocrParams)
{
ScannedImage = snapshot.Source;
TransformState = snapshot.TransformList.Count == 0 ? -1 : snapshot.TransformState;
Engine = ocrEngine;
OcrParams = ocrParams;
}
protected override RunInfo TesseractRunInfo(OcrParams ocrParams)
{
OcrMode mode = ocrParams.Mode;
string folder = mode == OcrMode.Fast || mode == OcrMode.Default ? "fast" : "best";
if (ocrParams.LanguageCode.Split('+').All(code => !File.Exists(Path.Combine(TesseractBasePath, folder, $"{code.ToLowerInvariant()}.traineddata"))))
{
// Use the other source if the selected one doesn't exist
folder = folder == "fast" ? "best" : "fast";
mode = folder == "fast" ? OcrMode.Fast : OcrMode.Best;
}
return(new RunInfo
{
Arguments = mode == OcrMode.Best ? "--oem 1" : mode == OcrMode.Legacy ? "--oem 0" : "",
DataPath = folder,
PrefixPath = folder
});
}
public async Task <OcrResult> QueueForeground(IOcrEngine ocrEngine, ScannedImage.Snapshot snapshot, string tempImageFilePath, OcrParams ocrParams, CancellationToken cancelToken)
{
OcrRequest req;
lock (this)
{
ocrEngine = ocrEngine ?? ocrManager.ActiveEngine ?? throw new ArgumentException("No OCR engine available");
ocrParams = ocrParams ?? ocrManager.DefaultParams;
var reqParams = new OcrRequestParams(snapshot, ocrEngine, ocrParams);
req = requestCache.GetOrSet(reqParams, () => new OcrRequest(reqParams));
// Fast path for cached results
if (req.Result != null)
{
SafeDelete(tempImageFilePath);
return(req.Result);
}
// Manage ownership of the provided temp file
if (req.TempImageFilePath == null)
{
req.TempImageFilePath = tempImageFilePath;
}
else
{
SafeDelete(tempImageFilePath);
}
// Increment the reference count
req.ForegroundCount += 1;
queueWaitHandle.Release();
}
// If no worker threads are running, start them
EnsureWorkerThreads();
// Wait for completion or cancellation
await Task.Factory.StartNew(() =>
{
try
{
WaitHandle.WaitAny(new[] { req.WaitHandle, cancelToken.WaitHandle });
}
catch (Exception e)
{
Log.ErrorException("Error in OcrRequestQueue.QueueForeground response task", e);
}
}, TaskCreationOptions.LongRunning);
lock (this)
{
// Decrement the reference count
req.ForegroundCount -= 1;
// If all requestors have cancelled and there's no result to cache, delete the request
DestroyRequest(req);
}
// If no requests are pending, stop the worker threads
EnsureWorkerThreads();
// May return null if cancelled
return(req.Result);
}
public bool HasCachedResult(IOcrEngine ocrEngine, ScannedImage.Snapshot snapshot, OcrParams ocrParams)
{
ocrEngine = ocrEngine ?? ocrManager.ActiveEngine ?? throw new ArgumentException("No OCR engine available");
ocrParams = ocrParams ?? ocrManager.DefaultParams;
var reqParams = new OcrRequestParams(snapshot, ocrEngine, ocrParams);
lock (this)
{
return(requestCache.ContainsKey(reqParams) && requestCache[reqParams].Result != null);
}
}
public void QueueBackground(ScannedImage.Snapshot snapshot, string tempImageFilePath, OcrParams ocrParams)
{
OcrRequest req;
CancellationTokenSource cts = new CancellationTokenSource();
lock (this)
{
var ocrEngine = ocrManager.ActiveEngine;
if (ocrEngine == null)
{
return;
}
ocrParams = ocrParams ?? ocrManager.DefaultParams;
var reqParams = new OcrRequestParams(snapshot, ocrEngine, ocrParams);
req = requestCache.GetOrSet(reqParams, () => new OcrRequest(reqParams));
// Fast path for cached results
if (req.Result != null)
{
return;
}
// Manage ownership of the provided temp file
if (req.TempImageFilePath == null)
{
req.TempImageFilePath = tempImageFilePath;
}
else
{
SafeDelete(tempImageFilePath);
}
// Increment the reference count
req.BackgroundCount += 1;
snapshot.Source.ThumbnailInvalidated += (sender, args) => cts.Cancel();
snapshot.Source.FullyDisposed += (sender, args) => cts.Cancel();
queueWaitHandle.Release();
}
// If no worker threads are running, start them
EnsureWorkerThreads();
var op = StartingOne();
Task.Factory.StartNew(() =>
{
try
{
WaitHandle.WaitAny(new[] { req.WaitHandle, cts.Token.WaitHandle, op.CancelToken.WaitHandle });
lock (this)
{
// Decrement the reference count
req.BackgroundCount -= 1;
// If all requestors have cancelled and there's no result to cache, delete the request
DestroyRequest(req);
}
FinishedOne();
// If no requests are pending, stop the worker threads
EnsureWorkerThreads();
}
catch (Exception e)
{
Log.ErrorException("Error in OcrRequestQueue.QueueBackground response task", e);
}
}, TaskCreationOptions.LongRunning);
}
public OcrResult ProcessImage(string imagePath, OcrParams ocrParams, CancellationToken cancelToken)
{
string tempHocrFilePath = Path.Combine(Paths.Temp, Path.GetRandomFileName());
string tempHocrFilePathWithExt = tempHocrFilePath + TesseractHocrExtension;
try
{
var runInfo = TesseractRunInfo(ocrParams);
var startInfo = new ProcessStartInfo
{
FileName = Path.Combine(TesseractBasePath, TesseractExePath),
Arguments = $"\"{imagePath}\" \"{tempHocrFilePath}\" -l {ocrParams.LanguageCode} {runInfo.Arguments} hocr",
UseShellExecute = false,
CreateNoWindow = true,
RedirectStandardOutput = true,
RedirectStandardError = true
};
if (runInfo.PrefixPath != null)
{
startInfo.EnvironmentVariables["TESSDATA_PREFIX"] = Path.Combine(TesseractBasePath, runInfo.PrefixPath);
}
if (runInfo.DataPath != null)
{
var tessdata = new DirectoryInfo(Path.Combine(TesseractBasePath, runInfo.DataPath));
EnsureHocrConfigExists(tessdata);
}
var tesseractProcess = Process.Start(startInfo);
if (tesseractProcess == null)
{
// Couldn't start tesseract for some reason
Log.Error("Couldn't start OCR process.");
return(null);
}
var timeout = (int)(appConfigManager.Config.OcrTimeoutInSeconds * 1000);
if (timeout == 0)
{
timeout = DEFAULT_TIMEOUT;
}
var stopwatch = Stopwatch.StartNew();
while (!tesseractProcess.WaitForExit(CHECK_INTERVAL))
{
if (stopwatch.ElapsedMilliseconds >= timeout || cancelToken.IsCancellationRequested)
{
if (stopwatch.ElapsedMilliseconds >= timeout)
{
Log.Error("OCR process timed out.");
}
try
{
tesseractProcess.Kill();
// Wait a bit to give the process time to release its file handles
Thread.Sleep(200);
}
catch (Exception e)
{
Log.ErrorException("Error killing OCR process", e);
}
return(null);
}
}
#if DEBUG && DEBUGTESS
Debug.WriteLine("Tesseract stopwatch: " + stopwatch.ElapsedMilliseconds);
var output = tesseractProcess.StandardOutput.ReadToEnd();
if (output.Length > 0)
{
Log.Error("Tesseract stdout: {0}", output);
}
output = tesseractProcess.StandardError.ReadToEnd();
if (output.Length > 0)
{
Log.Error("Tesseract stderr: {0}", output);
}
#endif
XDocument hocrDocument = XDocument.Load(tempHocrFilePathWithExt);
return(new OcrResult
{
PageBounds = hocrDocument.Descendants()
.Where(x => x.Attributes("class").Any(y => y.Value == "ocr_page"))
.Select(x => GetBounds(x.Attribute("title")))
.First(),
Elements = hocrDocument.Descendants()
.Where(x => x.Attributes("class").Any(y => y.Value == "ocrx_word"))
.Select(x => new OcrResultElement {
Text = x.Value, Bounds = GetBounds(x.Attribute("title"))
}),
RightToLeft = InstalledLanguages.Where(x => x.Code == ocrParams.LanguageCode).Select(x => x.RTL).FirstOrDefault()
});
}
catch (Exception e)
{
Log.ErrorException("Error running OCR", e);
return(null);
}
finally
{
try
{
File.Delete(tempHocrFilePathWithExt);
}
catch (Exception e)
{
Log.ErrorException("Error cleaning up OCR temp files", e);
}
}
}
protected virtual RunInfo TesseractRunInfo(OcrParams ocrParams) => new RunInfo
{
Arguments = "",
DataPath = "tessdata",
PrefixPath = ""
};
protected override RunInfo TesseractRunInfo(OcrParams ocrParams) => new RunInfo
{
Arguments = "",
DataPath = null,
PrefixPath = null
};