public Convolutional_Network(int inputHeight, int inputWidth)
{
Input_Layers_List ill = new Input_Layers_List(inputHeight, inputWidth);
layersList.Add(ill);
layersCount++;
}
public void ReadWeightsFromFile(string path)
{
layersList.Clear();
layersCount = 0;
string[] input = System.IO.File.ReadAllLines(@path);
int na = input.Count(i => i == "{");
List <string> s = input.ToList();
int p = 1;
for (int i = 0; i < na; i++)
{
if (s[p] == "0")
{
int inputH = int.Parse(s[p + 1]);
int inputW = int.Parse(s[p + 2]);
Input_Layers_List inp = new Input_Layers_List(inputH, inputW);
LayersList.Add(inp);
layersCount++;
p = s.IndexOf("{", p + 2);
}
else
{
List <List <List <double> > > weights = new List <List <List <double> > >();
int n = s.IndexOf("}", p);
var s1 = s.GetRange(p + 1, n - p - 1);
int nn = s1.Count(l => l == "<");
string type = s1[0];
int pp = 0;
for (int j = 0; j < nn; j++)
{
List <List <double> > w = new List <List <double> >();
pp = s1.IndexOf("<", pp) + 1;
int endIndex = s1.IndexOf(">", pp);
for (int str = pp; str < endIndex; str++)
{
List <double> temp = new List <double>();
string[] splitStr = s1[str].Split();
for (int l = 0; l < splitStr.Length - 1; l++)
{
temp.Add(double.Parse(splitStr[l]));
}
w.Add(temp);
}
weights.Add(w);
}
if (type == "2")
{
AddConvolutionalLayer(weights.Count / LayersList.Last().TablesCount, weights[0].Count, weights[0][0].Count);
LayersList.Last().SetWeigths(weights);
}
else
{
AddPullingLayer(weights[0].Count, weights[0][0].Count);
}
p = n + 1;
}
}
}
