/// <summary>
/// Args:
/// 0 - image path
/// 1 (optional, default: model.txt) - explicit model path
/// </summary>
/// <param name="args"></param>
static void Main(string[] args)
{
string resourcesPath = "../MovieClassifierLearner/MLSecondaryDataSet";
string modelPath = "model.txt";
//Load network from file
Network net = Network.LoadNetworkFromFile(modelPath);
//Load data set from dir
Console.WriteLine("Loading images...");
ImageLearningData[] data = ImageDataPreparer.PrepareImages(resourcesPath, 50, 25);
List <LabelID> ids = new List <LabelID>();
Console.WriteLine("Calculating averages...");
int a = 0;
foreach (ImageLearningData d in data)
{
Console.Write("\r" + a + "/" + data.Count());
net.PushInputValues(d.data.ToArray());
List <double> currentOutput = net.GetOutput();
if (ids.Exists(x => x.label == d.label))
{
LabelID i = ids.Find(x => x.label == d.label);
for (int j = 0; j < i.expectedOutput.Length; j++)
{
i.expectedOutput[j] += currentOutput[j];
}
i.hits++;
}
else
{
LabelID i = new LabelID();
i.label = d.label;
i.expectedOutput = currentOutput.ToArray();
i.hits = 1;
ids.Add(i);
}
a++;
}
foreach (LabelID i in ids)
{
Console.WriteLine(i.label + " is " + i.expectedOutput.MaxAt());
for (int j = 0; j < i.expectedOutput.Length; j++)
{
Console.Write((i.expectedOutput[j] / i.hits) + " ");
}
Console.WriteLine();
}
ids = ids.OrderBy(x => x.expectedOutput.MaxAt()).ToList();
Console.WriteLine("So the label.txt should look like this");
foreach (LabelID i in ids)
{
Console.WriteLine(i.label);
}
}
public static double[][][] Load(string path, int outputCount, int imageWidth, int imageHeight, string labelFilePath = "")
{
/*data:
* [0] -> Input Data to be evaluated
* [1] -> Expected Output Data
* [2] -> Test Input Data
* [3] -> Test Output Data*/
double[][][] finalData = new double[4][][];
List <double[]> inputData = new List <double[]>();
List <double[]> expectedOutputData = new List <double[]>();
List <double[]> testInputData = new List <double[]>();
List <double[]> testOutputData = new List <double[]>();
List <string> uniqueLabels = new List <string>();
if (labelFilePath != "")
{
string[] labels = File.ReadAllLines(labelFilePath);
uniqueLabels = labels.ToList();
}
{
//Ensure that ImageLearningData[] will be disposed after scope exit
ImageLearningData[] data = ImageDataPreparer.PrepareImages("Resources", imageWidth, imageHeight);
//Pack data into double Data table
for (int i = 0; i < data.Length; i++)
{
int labelIndex = uniqueLabels.IndexOf(data[i].label);
if (labelIndex == -1)
{
uniqueLabels.Add(data[i].label);
labelIndex = uniqueLabels.Count - 1;
}
//Assign expected output
double[] output = new double[outputCount];
output[labelIndex] = 1.0;
//Assign input values
double[] input = data[i].data.ToArray();
//Decide between test set and learning set
if (i % 10 == 0)
{
testInputData.Add(input);
testOutputData.Add(output);
}
else
{
inputData.Add(input);
expectedOutputData.Add(output);
}
}
}
//Shuffling
int[] numbers = new int[inputData.Count];
for (int i = 0; i < numbers.Length; i++)
{
numbers[i] = i;
}
Shuffler.Shuffle(numbers);
List <double[]> tmpInputData = new List <double[]>();
List <double[]> tmpOutputData = new List <double[]>();
for (int i = 0; i < numbers.Length; i++)
{
tmpInputData.Add(inputData[numbers[i]]);
tmpOutputData.Add(expectedOutputData[numbers[i]]);
}
inputData = tmpInputData;
expectedOutputData = tmpOutputData;
//Pack everything
finalData[0] = inputData.ToArray();
inputData.Clear();
finalData[1] = expectedOutputData.ToArray();
expectedOutputData.Clear();
finalData[2] = testInputData.ToArray();
testInputData.Clear();
finalData[3] = testOutputData.ToArray();
testOutputData.Clear();
//Output labels
File.WriteAllLines("labels.txt", uniqueLabels);
return(finalData);
}
