public static void Run()
{
var exporter = new PageXmlTextExporter(NearestNeighbourWordExtractor.Instance,
RecursiveXYCut.Instance,
UnsupervisedReadingOrderDetector.Instance,
scale: zoom);
using (var converter = new PdfImageConverter(pdfPath))
using (PdfDocument document = PdfDocument.Open(pdfPath))
{
var page = document.GetPage(pageNo);
var xml = exporter.Get(page);
File.WriteAllText(Path.ChangeExtension(pdfPath, pageNo + ".xml"), xml);
using (var bitmap = converter.GetPage(page.Number, zoom))
using (var graphics = Graphics.FromImage(bitmap))
{
// save pdf page as image
bitmap.Save(Path.ChangeExtension(pdfPath, pageNo + "_raw.png"));
// save empty image for LayoutEvalGUI
Bitmap blackAndWhite = new Bitmap(bitmap.Width, bitmap.Height, PixelFormat.Format8bppIndexed);
blackAndWhite.Save(Path.ChangeExtension(pdfPath, pageNo + "_bw_raw.png"));
}
}
}
public void WhenReadingOrder_ContainsReadingOrderXmlElements()
{
var pageXmlTextExporter = new PageXmlTextExporter(
DefaultWordExtractor.Instance,
RecursiveXYCut.Instance,
UnsupervisedReadingOrderDetector.Instance);
var xml = GetXml(pageXmlTextExporter);
Assert.Contains("<ReadingOrder>", xml);
Assert.Contains("</OrderedGroup>", xml);
}
public void NoPointsAreOnThePageBoundary()
{
var pageWidth = 100;
var pageHeight = 200;
var topLeftPagePoint = new PdfPoint(0, 0);
var bottomLeftPagePoint = new PdfPoint(0, pageHeight);
var bottomRightPagePoint = new PdfPoint(pageWidth, pageHeight);
var normalPoint = new PdfPoint(60, 60);
Assert.Equal("1,199", PageXmlTextExporter.PointToString(topLeftPagePoint, pageWidth, pageHeight));
Assert.Equal("1,1", PageXmlTextExporter.PointToString(bottomLeftPagePoint, pageWidth, pageHeight));
Assert.Equal("99,1", PageXmlTextExporter.PointToString(bottomRightPagePoint, pageWidth, pageHeight));
Assert.Equal($"60,140", PageXmlTextExporter.PointToString(normalPoint, pageWidth, pageHeight));
}
private static string GetXml(PageXmlTextExporter pageXmlTextExporter = null)
{
pageXmlTextExporter = pageXmlTextExporter ?? new PageXmlTextExporter(
DefaultWordExtractor.Instance,
RecursiveXYCut.Instance,
UnsupervisedReadingOrderDetector.Instance);
string xml;
using (var document = PdfDocument.Open(GetFilename()))
{
var page = document.GetPage(1);
xml = pageXmlTextExporter.Get(page);
}
return(xml);
}
