private static void InteractiveDemo()
{
Console.WriteLine("Entering interactive mode...");
Console.WriteLine("Which feature do you want to use? (Input an integer.)");
Console.WriteLine("[1] Color moment");
Console.WriteLine("[2 - 6] RGB Histgram, with dim 2,3,4,5,6 in each channel");
Console.WriteLine("[7 - 11] HSL Histgram, with dim 2,3,4,5,6 in each channel");
Console.WriteLine("[12] 4x4 Binary density");
var strategy = int.Parse(Console.ReadLine() ?? throw new Exception());
Console.WriteLine("Building R Tree...");
const int minEntry = 12;
const int maxEntry = 30;
var featurePath = StrategyMap[strategy].Item1;
var featureDim = StrategyMap[strategy].Item2;
var tree = new RTree <string>(maxEntry, minEntry);
var featureLines = File.ReadAllLines(featurePath);
var points = featureLines.Select(s =>
{
var slices = s.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
return(new Point(featureDim, slices.Select(double.Parse).ToList()));
}).ToList();
var imageNames = File.ReadAllLines(ImageFilePath);
for (var i = 0; i < NumImages; i++)
{
tree.AddRecord(new Rectangle(featureDim, points[i], points[i]), imageNames[i]);
}
var imagePointMap = new Dictionary <string, Point>();
for (var i = 0; i < NumImages; i++)
{
imagePointMap.Add(imageNames[i], points[i]);
}
while (true)
{
Console.WriteLine(
"Please choose a image by file name (e.g. \"n01613177_3.JPEG\", quotation mark is not needed) :");
var imageName = Console.ReadLine();
if (imageName == null)
{
throw new Exception();
}
Console.WriteLine("How many items do you want to retrievel?");
var topK = int.Parse(Console.ReadLine() ?? throw new Exception());
var results = tree.GetKNearestItems(imagePointMap[imageName], topK);
results.ForEach(res =>
{
Console.WriteLine($"{res} {string.Join(" ", imagePointMap[res].Coordinate)}");
});
}
}
private static int GetNumDiskAccess(int rtreeSize, int strategy, int minEntry, int maxEntry)
{
var featurePath = StrategyMap[strategy].Item1;
var featureDim = StrategyMap[strategy].Item2;
var nums = Enumerable.Range(0, NumImages).ToList();
nums.Shuffle();
var tree = new RTree <string>(maxEntry, minEntry);
var featureLines = File.ReadAllLines(featurePath);
var points = featureLines.Select(s =>
{
var slices = s.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
return(new Point(featureDim, slices.Select(double.Parse).ToList()));
}).ToList();
var imageNames = File.ReadAllLines(ImageFilePath);
for (var i = 0; i < rtreeSize; i++)
{
tree.AddRecord(new Rectangle(featureDim, points[nums[i]], points[nums[i]]), imageNames[nums[i]]);
}
var mbr = tree.GetRootMbr();
var ratio = Math.Pow(1d / 2, 1d / featureDim);
const int queryTimes = 1024;
var countSum = 0;
for (var i = 0; i < queryTimes; i++)
{
var queryMinBound = Enumerable.Repeat(0d, featureDim).ToList();
var queryMaxBound = Enumerable.Repeat(0d, featureDim).ToList();
for (var dim = 0; dim < featureDim; dim++)
{
var length = mbr.MaxBoundries[dim] - mbr.MinBoundries[dim];
var sideLength = length * ratio;
var start = mbr.MinBoundries[dim] + new Random(DateTime.Now.Millisecond).NextDouble() *
(mbr.MaxBoundries[dim] - sideLength - mbr.MinBoundries[dim]);
var end = start + sideLength;
queryMinBound[dim] = start;
queryMaxBound[dim] = end;
}
var queryRect = new Rectangle(featureDim, queryMinBound, queryMaxBound);
tree.GetContainedItems(queryRect, out var count);
countSum += count;
}
return(countSum / queryTimes);
}
private static double GetAccuracy(int strategy, int numTopK)
{
const int minEntry = 12;
const int maxEntry = 30;
var featurePath = StrategyMap[strategy].Item1;
var featureDim = StrategyMap[strategy].Item2;
var tree = new RTree <string>(maxEntry, minEntry);
var featureLines = File.ReadAllLines(featurePath);
var points = featureLines.Select(s =>
{
var slices = s.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
return(new Point(featureDim, slices.Select(double.Parse).ToList()));
}).ToList();
var imageNames = File.ReadAllLines(ImageFilePath);
for (var i = 0; i < NumImages; i++)
{
tree.AddRecord(new Rectangle(featureDim, points[i], points[i]), imageNames[i]);
}
var totalAccuracy = 0d;
for (var i = 0; i < NumImages; i++)
{
var results = tree.GetKNearestItems(points[i], numTopK);
results.Remove(imageNames[i]);
var total = results.Count;
var count = 0;
foreach (var result in results)
{
if (result.Split('_')[0] == imageNames[i].Split('_')[0])
{
count++;
}
}
var accuracy = (double)count / total;
totalAccuracy += accuracy;
}
return(totalAccuracy / NumImages);
}
private static int GetSplitCountOfBuilding(int rtreeSize, int strategy, int minEntry, int maxEntry)
{
var tree = new RTree <string>(maxEntry, minEntry);
var featurePath = StrategyMap[strategy].Item1;
var featureDim = StrategyMap[strategy].Item2;
var featureLines = File.ReadAllLines(featurePath);
var points = featureLines.Select(s =>
{
var slices = s.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
return(new Point(featureDim, slices.Select(double.Parse).ToList()));
}).ToList();
var imageNames = File.ReadAllLines(ImageFilePath);
for (var i = 0; i < NumImages; i++)
{
tree.AddRecord(new Rectangle(featureDim, points[i], points[i]), imageNames[i]);
}
return(tree.SplitCount);
}
private static double GetRecall(int strategy)
{
const int minEntry = 12;
const int maxEntry = 30;
var tree = new RTree <string>(maxEntry, minEntry);
var featurePath = StrategyMap[strategy].Item1;
var featureDim = StrategyMap[strategy].Item2;
var featureLines = File.ReadAllLines(featurePath);
var points = featureLines.Select(s =>
{
var slices = s.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
return(new Point(featureDim, slices.Select(double.Parse).ToList()));
}).ToList();
var imageNames = File.ReadAllLines(ImageFilePath);
for (var i = 0; i < NumImages; i++)
{
tree.AddRecord(new Rectangle(featureDim, points[i], points[i]), imageNames[i]);
}
var dict = new Dictionary <string, int>();
for (var i = 0; i < NumImages; i++)
{
var category = imageNames[i].Split('_')[0];
if (dict.Keys.Contains(category))
{
dict[category]++;
}
else
{
dict[category] = 1;
}
}
var totalRecall = 0d;
for (var i = 0; i < NumImages; i++)
{
var category = imageNames[i].Split('_')[0];
var results = tree.GetKNearestItems(points[i], dict[category]);
var count = 0;
foreach (var result in results)
{
if (result.Split('_')[0] == category)
{
count++;
}
}
var recall = (double)count / dict[category];
totalRecall += recall;
if (i % 100 == 0)
{
Console.Error.WriteLine(i);
}
}
return(totalRecall / NumImages);
}