public void RunRankingModels(Picture picture, Picture.Answer query)
{
for (int datasetIndex = 0; datasetIndex < PreprocessedDatasets.Length; datasetIndex++)
{
for (int rankModIndex = 0; rankModIndex < RankingModels.Count; rankModIndex++)
{
RunSingleRankingModel(picture, query, datasetIndex, rankModIndex);
}
}
}
public int RunSingleRankingModel(Picture picture, Picture.Answer query, int datasetIndex, int rankModIndex)
{
var searchResults = RankingModels[rankModIndex].Sort(query, PreprocessedDatasets[datasetIndex]);
var rank = searchResults.FindIndex(i => Path.GetFileName(i.SourceFile) == Path.GetFileName(picture.Filename));
Histograms[datasetIndex, rankModIndex, rank]++;
OnOneRankingFinished(datasetIndex, rankModIndex, query, rank);
return(rank);
}
public override List <ImageWarehouse.Image> Sort(Picture.Answer query, ConcurrentBag <ImageWarehouse.Image> imageDataset)
{
var outputList = new List <ImageWarehouse.Image>(imageDataset.Count);
foreach (var img in imageDataset)
{
outputList.Add(img);
}
DebugSave(outputList);
return(outputList);
}
public abstract List <Image> Sort(Picture.Answer query, ConcurrentBag <Image> imageDataset);
public override List <ImageWarehouse.Image> Sort(Picture.Answer Query, ConcurrentBag <ImageWarehouse.Image> imageDataset)
{
var results = new ConcurrentBag <(double, ImageWarehouse.Image)>();
Parallel.ForEach(imageDataset, img => {
double totalDist = 0;
foreach (var circle in Query.Circles)
{
int relativeX = (int)(circle.X / ((double)img.Width / img.PreprocessedImage.GetLength(0)));
int relativeY = (int)(circle.Y / ((double)img.Height / img.PreprocessedImage.GetLength(1)));
if (relativeX >= img.PreprocessedImage.GetLength(0) || relativeY >= img.PreprocessedImage.GetLength(1))
{
// this happens, but why?? how could this error happen?
// are the data mixed up in some way?
return;
}
var targetPoint = targetFinder.FindTarget(circle, img);
var targetColor = img.PreprocessedImage[targetPoint.X, targetPoint.Y];
var distX = relativeX - targetPoint.X;
var distY = relativeY - targetPoint.Y;
var lastRmDist = Distance(circle.Color, targetColor, distX, distY);
totalDist += lastRmDist;
}
double avgDist = totalDist / Query.Circles.Length;
if (double.IsNaN(avgDist))
{
Console.WriteLine("nananananana");
}
if (double.IsInfinity(avgDist))
{
Console.WriteLine("to infinityyyy");
}
results.Add((avgDist, img));
});
var orderedResults = from r in results
orderby r.Item1
select r;
var orderedTop100 = orderedResults.Take(100);
// zkuste k tomu take vypisovat pro kazdy ranking ciselnou hodnotu H, kterou ziskate z prvnich 100 prvku serazenych vysledku(pro velkou DB)
// H = -k / (sum_{ i = 1..100} ln(x_i / x_100)), kde x je distance podle ktere obrazky tridite
//
// pokud budeme mit tuto hodnotu pro RANK a pak pro RANK bez odebraneho kolecka, tak muzeme sledovat ve 2D, jestli tvori dobra a spatna kolecka klastry nebo ne
int k = results.Count; // ?? je to správne
// (sum_{ i = 1..100} ln(x_i / x_100)), kde x je distance podle ktere obrazky tridite
var lastDistance = orderedTop100.Last().Item1;
var valueHEnum = from r in orderedTop100 select r.Item1 == 0 || lastDistance == 0 ? 0 : Math.Log(r.Item1 / lastDistance);
double valueHSum = valueHEnum.Sum();
return((from r in results orderby r.Item1 select r.Item2).ToList());
}
private void ImageWarehouse_OnOneRankingFinished(int datasetIndex, int rankModelIndex, Picture.Answer query, int rank)
{
if (progressBar2.IsHandleCreated)
{
progressBar2.BeginInvoke((MethodInvoker)(() =>
{
progressBar2.Value++;
}));
}
}
