/// <summary>
/// Refreshes the node metadata in all trees in the forest.
/// </summary>
public void RefreshMetadata()
{
_leafCount = 0;
_levelCount = 0;
_testCounts = new Dictionary <string, int>();
for (byte i = 0; i < _numTrees; i++)
{
DecisionTree <T, D> tree = _trees[i];
_leafCount = tree.SetTreeLabel(i, _leafCount);
_levelCount = Math.Max(_levelCount, tree.LevelCount);
Dictionary <string, int> testCounts = tree.TestCounts;
foreach (string key in testCounts.Keys)
{
if (!_testCounts.ContainsKey(key))
{
_testCounts[key] = 0;
}
_testCounts[key] += testCounts[key];
}
}
}
/// <summary>
/// Trains a decision forest from <paramref name="splits"/> based on the provided parameters using the depth first algorithm.
/// </summary>
/// <param name="numTrees">Number of trees in the forest</param>
/// <param name="splits">Data splits to use when training the tree.</param>
/// <param name="factory">The feature factory</param>
/// <param name="numFeatures">The number of features to try for each node</param>
/// <param name="numThresholds">The number of thresholds to try for each node</param>
/// <param name="labelNames">The names for the labels</param>
/// <param name="labelWeights">An array of weights for each label</param>
/// <returns>The trained forest</returns>
public static DecisionForest <T, D> ComputeDepthFirst(
int numTrees,
List <T>[] splits,
IFeatureFactory <T, D> factory,
int numFeatures,
int numThresholds,
string[] labelNames,
float[] labelWeights
)
{
int numLabels = labelNames.Length;
DecisionTree <T, D>[] trees = new DecisionTree <T, D> [numTrees];
int count = 0;
var indices = Enumerable.Range(0, numTrees).Select(o => (byte)o);
if (splits[0][0] is IComparable <T> )
{
foreach (var split in splits)
{
split.Sort();
}
}
foreach (var i in indices)
{
int split = i % splits.Length;
UpdateManager.WriteLine(string.Format("Training tree {0} of {1}...", i + 1, numTrees));
trees[i] = DecisionTree <T, D> .ComputeDepthFirst(splits[split], factory, numFeatures, numThresholds, numLabels, labelWeights);
trees[i].LabelCount = labelNames.Length;
count = trees[i].SetTreeLabel(i, count);
UpdateManager.WriteLine("\ndone");
}
;
UpdateManager.WriteLine("Training complete");
return(new DecisionForest <T, D>(trees, labelNames));
}
/// <summary>
/// Classifies each point from <paramref name="image"/> and trackes which nodes it visits.
/// </summary>
/// <param name="tree">The tree used for the computation</param>
/// <param name="image">Image to add to the tree</param>
/// <param name="mode">Mode to use when sampling the image</param>
public static void Fill <T>(this DecisionTree <ImageDataPoint <T>, T[]> tree, LabeledImage <T> image, BackgroundSampleMode mode)
{
List <ImageDataPoint <T> > points = image.CreateAllDataPoints(mode);
tree.Fill(points);
}
/// <summary>
/// Classifies each pixel of <paramref name="image"/> and produces a corresponding <see cref="T:LabelImage" />. The maximum likelihood label
/// is chosen at each pixel.
/// </summary>
/// <param name="tree">The tree used for the computation</param>
/// <param name="image">Image to classify</param>
/// <returns>A label image with all of the classifications</returns>
public static LabelImage Classify <T>(this DecisionTree <ImageDataPoint <T>, T[]> tree, IMultichannelImage <T> image)
{
return(tree.ClassifySoft(image).ToLabelImage());
}
