private static classificacao ClassifySingleImage(MLContext mlContext, string imagePath, string outputModelLocation, ITransformer model)
{
var imageData = new ImageData()
{
ImagePath = imagePath
};
// Make prediction function (input = ImageData, output = ImagePrediction)
var predictor = mlContext.Model.CreatePredictionEngine <ImageData, ImagePrediction>(model);
var prediction = predictor.Predict(imageData);
classificacao c = new classificacao();
c.tipo = prediction.PredictedLabelValue;
c.porcentagem = prediction.Score.Max();
return(c);
}
public List <classificacao> analisar(string path, int deltacor, int instancia)
{
List <classificacao> listaclassific = new List <classificacao>();
whidthoriginal = new Bitmap(path).Width;
heightoriginal = new Bitmap(path).Height;
Bitmap bitmap = new Bitmap(new Bitmap(path), 250, 250);
List <Point> listap = new List <Point>();
//int deltanominal = 10;
int deltanominal = deltacor;
Bitmap b = new Bitmap(bitmap);
for (int i = 0; i < 250; i++)
{
for (int i2 = 0; i2 < 250; i2++)
{
double media = (b.GetPixel(i, i2).R + b.GetPixel(i, i2).G + b.GetPixel(i, i2).B) / 3;
int delta = Convert.ToInt32(media - deltanominal);
if (delta < -1)
{
delta = delta * -1;
}
if (delta <= 30)
{
listap.Add(new Point(i, i2));
}
}
}
List <pointtipe> lpt = new List <pointtipe>();
List <pointtipe> lptwatch = new List <pointtipe>();
List <Point> listap2 = new List <Point>();
for (int i = 0; i < 250; i += 3)
{
for (int i2 = 0; i2 < 250; i2 += 3)
{
if (listap.FirstOrDefault(x => x.X == i && x.Y == i2) != null)
{
listap2.Add(listap.FirstOrDefault(x => x.X == i && x.Y == i2));
}
}
}
int deltad = 10;
foreach (var item in listap2)
{
lpt.Add(new pointtipe(item, -1));
}
int classeatual = 0;
for (int i = 0; i < lpt.Count(); i++)
{
if (lpt[i].classe == -1)
{
lpt[i].classe = classeatual;
lptwatch.Add(lpt[i]);
do
{
for (int i2 = 0; i2 < lptwatch.Count(); i2++)
{
foreach (var item2 in lpt.Where(x => x.classe == -1 && x.ponto != lptwatch[i2].ponto))
{
double distancia;
double n1 = (lptwatch[i2].ponto.X - item2.ponto.X) * (lptwatch[i2].ponto.X - item2.ponto.X);
double n2 = (lptwatch[i2].ponto.Y - item2.ponto.Y) * (lptwatch[i2].ponto.Y - item2.ponto.Y);
distancia = Math.Sqrt(n1 + n2);
if (distancia <= deltad && distancia >= deltad - 1)
{
item2.classe = classeatual;
lptwatch.Add(item2);
}
else if (distancia <= deltad && distancia < deltad - 1)
{
item2.classe = classeatual;
}
}
lptwatch.Remove(lptwatch.First());
}
} while (lptwatch.Count() > 0);
classeatual++;
}
}
Bitmap b2 = new Bitmap(250, 250);
List <pointtipe> listapontosreais = new List <pointtipe>();
foreach (var item in lpt)
{
if (lpt.Where(x => x.classe == item.classe).ToList().Count() > 7)
{
listapontosreais.Add(item);
}
}
List <List <pointtipe> > llp = new List <List <pointtipe> >();
for (int i = 0; i < lpt.ToList().Count(); i++)
{
if (listapontosreais.Where(x => x.classe == i).ToList().Count() > 0)
{
llp.Add(listapontosreais.Where(x => x.classe == i).ToList());
}
}
int direc = 0;
do
{
try
{
Directory.CreateDirectory($"imagensgrid{instancia}");
listaclassific = new List <classificacao>();
int n = 0;
foreach (var item in llp)
{
pointtipe ltb = item.OrderBy(z => z.ponto.X).First();
ltb.ponto = new Point(ltb.ponto.X, item.OrderBy(z => z.ponto.Y).First().ponto.Y);
pointtipe rbb = item.OrderByDescending(z => z.ponto.X).First();
rbb.ponto = new Point(rbb.ponto.X, item.OrderByDescending(z => z.ponto.Y).First().ponto.Y);
if (ltb.ponto.X != rbb.ponto.X && ltb.ponto.Y != rbb.ponto.Y)
{
Bitmap bitb = new Bitmap((rbb.ponto.X - ltb.ponto.X), (rbb.ponto.Y - ltb.ponto.Y));
for (int i = ltb.ponto.X; i < rbb.ponto.X; i++)
{
for (int i2 = ltb.ponto.Y; i2 < rbb.ponto.Y; i2++)
{
bitb.SetPixel(i - ltb.ponto.X, i2 - ltb.ponto.Y, bitmap.GetPixel(i, i2));
}
}
bitb.Save($"imagensgrid{instancia}/imgdetect{n}.png");
classificacao c = new classificacao();
double mx = (double)whidthoriginal / (double)250;
double my = (double)heightoriginal / (double)250;
c.x = Convert.ToInt32(Math.Round((double)ltb.ponto.X * mx));
c.y = Convert.ToInt32(Math.Round((double)ltb.ponto.Y * my));
c.width = Convert.ToInt32(Math.Round((double)(rbb.ponto.X - ltb.ponto.X) * mx));
c.height = Convert.ToInt32(Math.Round((double)(rbb.ponto.Y - ltb.ponto.Y) * my));
c.path = $"imagensgrid{instancia}/imgdetect{n}.png";
listaclassific.Add(c);
n++;
}
}
break;
}
catch (Exception)
{ }
} while (true);
foreach (var item in listaclassific)
{
classificacao s = ClassifySingleImage(mlContext, Path.GetFullPath(item.path), _outputImageClassifierZip, model);
item.porcentagem = s.porcentagem;
item.tipo = s.tipo;
}
return(listaclassific);
}