public async Task <List <TextPage> > DecodeHocr(string text)
{
return(await Task.Run(async() =>
{
var divName = XName.Get("div");
List <TextPage> textPages = new List <TextPage>();
XDocument textXml = XDocument.Parse(text);
IEnumerable <XElement> pages = textXml.Descendants(divName)
.Where(x => (string)x.Attribute("class") == "ocr_page");
foreach (XElement page in pages)
{
TextPage textPage = new TextPage();
XAttribute coords = page.Attribute("title");
if (coords != null)
{
string[] coordlist = coords.Value.Split(' ');
textPage.X = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[3]));
textPage.Y = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[4]));
textPage.Width = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[5])) -
Convert.ToInt32(HelperOcr.GetNumbers(coordlist[3]));
textPage.Height = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[6])) -
Convert.ToInt32(HelperOcr.GetNumbers(coordlist[4]));
}
XAttribute id = page.Attribute("id");
if (id != null)
{
textPage.Id = id.Value;
}
List <TextParagraph> textParagraphs = await Task.Run(() => new DecodeParagraphs().Decode(page));
textPage.Paragraphs.AddRange(textParagraphs);
textPages.Add(textPage);
}
return textPages;
}));
}
public List <TextPage> Decode(string text)
{
List <TextPage> textPages = new List <TextPage>();
XDocument textXml = XDocument.Parse(text);
IEnumerable <XElement> pages = textXml.Descendants(_divName)
.Where(x => (string)x.Attribute("class") == "ocr_page");
foreach (XElement page in pages)
{
TextPage textPage = new TextPage();
XAttribute coords = page.Attribute("title");
if (coords != null)
{
string[] coordlist = coords.Value.Split(' ');
textPage.X = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[3]));
textPage.Y = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[4]));
textPage.Width = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[5])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[3]));
textPage.Height = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[6])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[4]));
}
XAttribute id = page.Attribute("id");
if (id != null)
{
textPage.Id = id.Value;
}
List <TextParagraph> textParagraphs = new DecodeParagraphs().Decode(page);
textPage.Paragraphs.AddRange(textParagraphs);
textPages.Add(textPage);
}
SerializePagesList.Serialize(textPages);
return(textPages);
}
public List <TextWord> Decode(XElement line)
{
List <TextWord> textWords = new List <TextWord>();
IEnumerable <XElement> words = line.Descendants(_spanName);
foreach (XElement word in words)
{
TextWord textWord = new TextWord();
XAttribute coords = word.Attribute("title");
if (coords != null)
{
var coordlist = coords.Value.Split(' ');
textWord.X = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[1]));
textWord.Y = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[2]));
textWord.Width = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[3])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[1]));
textWord.Height = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[4])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[2]));
}
XAttribute id = word.Attribute("id");
if (word.Elements(XName.Get("strong")).FirstOrDefault(d => !d.IsEmpty) != null)
{
textWord.Bold = true;
}
;
if (id != null)
{
textWord.id = id.Value;
}
textWord.Word = word.Value;
textWords.Add(textWord);
}
return(textWords);
}
public List <TextParagraph> Decode(XElement element)
{
List <TextParagraph> textParagraphs = new List <TextParagraph>();
IEnumerable <XElement> paragraphs = element.Descendants(_paraName)
.Where(x => (string)x.Attribute("class") == "ocr_par");
foreach (XElement paragraph in paragraphs)
{
TextParagraph textParagraph = new TextParagraph();
XAttribute coords = paragraph.Attribute("title");
if (coords != null)
{
string[] coordlist = coords.Value.Split(' ');
textParagraph.X = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[1]));
textParagraph.Y = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[2]));
textParagraph.Width = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[3])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[1]));
textParagraph.Height = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[4])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[2]));
}
XAttribute id = paragraph.Attribute("id");
if (id != null)
{
textParagraph.id = id.Value;
}
List <TextLine> textLines = new DecodeLines().Decode(paragraph);
textParagraph.Lines.AddRange(textLines);
textParagraphs.Add(textParagraph);
}
return(textParagraphs);
}
public List <TextLine> Decode(XElement paragraph)
{
List <TextLine> textLines = new List <TextLine>();
IEnumerable <XElement> lines = paragraph.Descendants(_spanName)
.Where(x => (string)x.Attribute("class") == "ocr_line");
foreach (XElement line in lines)
{
TextLine textLine = new TextLine();
XAttribute coords = line.Attribute("title");
if (coords != null)
{
string[] coordlist = coords.Value.Split(' ');
textLine.X = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[1]));
textLine.Y = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[2]));
textLine.Width = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[3])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[1]));
textLine.Height = Convert.ToInt32(HelperOcr.GetNumbers(coordlist[4])) - Convert.ToInt32(HelperOcr.GetNumbers(coordlist[2]));
}
XAttribute id = line.Attribute("id");
if (id != null)
{
textLine.id = id.Value;
}
List <TextWord> textWords = new DecodeWords().Decode(line);
textLine.Words.AddRange(textWords);
textLines.Add(textLine);
}
return(textLines);
}
