public async Task <OcrHWResult> GetText(Stream stream, string url = null)
{
var visionResult = await(stream != null ? visionClient.RecognizeTextAsync(stream, "en", false) : visionClient.RecognizeTextAsync(url, "en", false));
var lines = visionResult.Regions.SelectMany(r => r.Lines).Select(l =>
new lineResult()
{
boundingBox = ConvertBoundingBox(l.BoundingBox),
words = l.Words.Select(w => new WordResult()
{
boundingBox = ConvertBoundingBox(w.BoundingBox),
text = w.Text
}).ToArray()
}
);
var result = new OcrHWResult()
{
lines = lines.ToArray()
};
return(result);
}
public OcrHWResult Concat(OcrHWResult result)
{
var newResult = new OcrHWResult();
newResult.lines = lines.Concat(result.lines).ToArray();
newResult.Tags = Tags.Concat(result.Tags).ToArray();
newResult.Width = Width + result.Width;
newResult.Height = Height + result.Height;
return(newResult);
}
public async Task <OcrHWResult> GetVision(Stream stream, string url = null)
{
var features = new[] { VisualFeature.Tags, VisualFeature.ImageType, VisualFeature.Description, VisualFeature.Adult };
var visionResult = stream != null ?
await visionClient.AnalyzeImageAsync(stream, features)
: await visionClient.AnalyzeImageAsync(url, features);
List <lineResult> lines = new List <lineResult>();
lines.AddRange(visionResult.Description.Captions.Select(c => new lineResult()
{
words = c.Text.Split(' ').Select(w => new WordResult()
{
text = w
}).ToArray()
}
));
lines.Add(new lineResult()
{
words = visionResult.Tags.Select(t => new WordResult()
{
text = t.Name
}).ToArray()
});
var result = new OcrHWResult()
{
lines = lines.ToArray(),
AdultScore = visionResult.Adult.AdultScore,
RacyScore = visionResult.Adult.RacyScore,
Tags = visionResult.Tags.Select(t => t.Name),
Width = visionResult.Metadata.Width,
Height = visionResult.Metadata.Height,
};
return(result);
}
public static OcrHWResult FixLinesAndWordOrder(OcrHWResult result)
{
var allWords = result.lines.SelectMany(l => l.words);
var avgWordHeight = (float)allWords.Average(w => w.Height);
int lineTol = (int)(avgWordHeight * .4);
var wordGroups = GetClusters(allWords, l => l.CenterY, lineTol);
var foo = wordGroups.OrderBy(lg => lg.First().CenterY);
Console.WriteLine();
Console.WriteLine("text:");
foreach (var lg in foo)
{
//Console.WriteLine("Y: " + lg.First().CenterY);
foreach (var l in lg.OrderBy(l => l.StartX))
{
Console.Write(" " + l.text);
}
Console.WriteLine();
}
return(result);
}
public IEnumerable <WordResult> AddPage(OcrHWResult hw, string imageUrl)
{
// page
metadata.WriteLine($" <div class='ocr_page' id='page_{pageCount}' title='image \"{imageUrl}\"; bbox 0 0 {hw.Width} {hw.Height}; ppageno {pageCount}'>");
metadata.WriteLine($" <div class='ocr_carea' id='block_{pageCount}_1'");
var allwords = new List <WordResult>();
int li = 0;
int wi = 0;
foreach (var line in hw.lines /*.SortLines(hw.lines)*/)
{
metadata.WriteLine($" <span class='ocr_line' id='line_{pageCount}_{li}' title='baseline -0.002 -5; x_size 30; x_descenders 6; x_ascenders 6'>");
var words = line.words.FirstOrDefault()?.boundingBox == null ? line.words : line.words.OrderBy(l => l.boundingBox[0]).ToArray();
foreach (var word in words)
{
var bbox = word.boundingBox != null && word.boundingBox.Length == 8 ? $"bbox {word.boundingBox[0]} {word.boundingBox[1]} {word.boundingBox[4]} {word.boundingBox[5]}" : "";
metadata.WriteLine($" <span class='ocrx_word' id='word_{pageCount}_{li}_{wi}' title='{bbox}'>{word.text}</span>");
text.WriteLine(word.text);
wi++;
allwords.Add(word);
}
li++;
metadata.WriteLine(" </span>"); // line
}
metadata.WriteLine(" </div>"); // reading area
metadata.WriteLine(" </div>"); // page
pageCount++;
return(allwords);
}
public async Task <OcrHWResult> GetHandwritingText(Stream stream, string url = null)
{
var client = new HttpClient();
client.DefaultRequestHeaders.Add("Ocp-Apim-Subscription-Key", "92e5723b229149519f717c1e1ae81443");
var uri = "https://westus.api.cognitive.microsoft.com/vision/v1.0/recognizeText?handwriting=true";
HttpResponseMessage response;
// Request body
if (stream != null)
{
using (var content = new StreamContent(stream))
{
content.Headers.ContentType = new MediaTypeHeaderValue("application/octet-stream");
response = await client.PostAsync(uri, content);
}
}
else
{
var json = JsonConvert.SerializeObject(new { url = url });
using (var content = new StringContent(json))
{
content.Headers.ContentType = new MediaTypeHeaderValue("application/json");
response = await client.PostAsync(uri, content);
}
}
OcrHWResult result = null;
IEnumerable <string> opLocation;
if (!response.IsSuccessStatusCode)
{
var err = await response.Content.ReadAsStringAsync();
response.EnsureSuccessStatusCode();
}
if (response.Headers.TryGetValues("Operation-Location", out opLocation))
{
while (true)
{
response = await client.GetAsync(opLocation.First());
var txt = await response.Content.ReadAsStringAsync();
var status = JsonConvert.DeserializeObject <AsyncStatusResult>(txt);
if (status.status == "Running" || status.status == "NotStarted")
{
await Task.Delay(TimeSpan.FromMilliseconds(500));
}
else
{
result = status.recognitionResult;
break;
}
}
}
return(result);
}
