public void Test_Input2D_NullInput()
{
Data2D data = null;
Input2DLayer inp = new Input2DLayer();
inp.SetInput(data);
}
public void Test_AvgPool2D_Null_Input()
{
Data2D data = null;
AvgPool2DLayer pool = new AvgPool2DLayer(1, 1, 1, 1, 2, 2);
pool.SetInput(data);
}
public void Test_Reshape2D_WrongSIzes()
{
Data2D data = new Data2D(2, 3, 5, 2);
Reshape2DLayer res = new Reshape2DLayer(1, 2, 1, 4);
res.SetInput(data);
}
public void Test_MaxPool1D_Execute()
{
// Initialize data.
Data2D data = new Data2D(1, 3, 2, 1);
data[0, 0, 0, 0] = 1;
data[0, 1, 0, 0] = 2;
data[0, 2, 0, 0] = 0;
data[0, 0, 1, 0] = 3;
data[0, 1, 1, 0] = 4;
data[0, 2, 1, 0] = 0;
MaxPool1DLayer pool = new MaxPool1DLayer(0, 1, 2);
pool.SetInput(data);
pool.Execute();
Data2D output = pool.GetOutput() as Data2D;
// Checking sizes
Dimension dim = output.GetDimension();
Assert.AreEqual(dim.b, 1);
Assert.AreEqual(dim.c, 2);
Assert.AreEqual(dim.h, 1);
Assert.AreEqual(dim.w, 2);
// Checking calculation
Assert.AreEqual(output[0, 0, 0, 0], 2.0, 0.0000001);
Assert.AreEqual(output[0, 1, 0, 0], 2.0, 0.0000001);
Assert.AreEqual(output[0, 0, 1, 0], 4.0, 0.0000001);
Assert.AreEqual(output[0, 1, 1, 0], 4.0, 0.0000001);
}
public void Test_ReLu_Execute()
{
relu = new ReLuLayer();
Data2D data = new Data2D(2, 3, 1, 1);
data[0, 0, 0, 0] = 4;
data[0, 1, 0, 0] = 2;
data[0, 2, 0, 0] = -2;
data[1, 0, 0, 0] = 3;
data[1, 1, 0, 0] = -1;
data[1, 2, 0, 0] = -3;
relu.SetInput(data);
relu.Execute();
Data2D output = relu.GetOutput() as Data2D;
Assert.AreEqual(output[0, 0, 0, 0], 4.0, 0.00000001);
Assert.AreEqual(output[0, 1, 0, 0], 2.0, 0.00000001);
Assert.AreEqual(output[0, 2, 0, 0], 0.0, 0.00000001);
Assert.AreEqual(output[1, 0, 0, 0], 3.0, 0.00000001);
Assert.AreEqual(output[1, 1, 0, 0], 0.0, 0.00000001);
Assert.AreEqual(output[1, 2, 0, 0], 0.0, 0.00000001);
}
public void Test_Cropping2D_Negative_Trim()
{
Data2D data = new Data2D(8, 4, 3, 5);
Cropping2DLayer crop = new Cropping2DLayer(4, -5, -1, 1);
crop.SetInput(data);
}
public double[] Evaluate(Bitmap img)
{
if (model == null)
{
var reader = new ReaderKerasModel(cnn_nn);
model = reader.GetSequentialExecutor();
}
var array = new Data2D(28, 28, 1, 1);
for (int i = 0; i < img.Height; i++)
{
for (int j = 0; j < img.Width; j++)
{
Color pixel = img.GetPixel(j, i);
double value = 255 - pixel.A;
value = value / 255;
array[i, j, 0, 0] = value;
}
}
var result = model.ExecuteNetwork(array) as Data2D;
double[] toreturn = new double[10];
for (int i = 0; i < 10; i++)
{
toreturn[i] = result[0, 0, i, 0];
}
return(toreturn);
}
public void Test_GlobalMaxPool1D_Null_Input()
{
Data2D data = null;
GlobalMaxPool1DLayer pool = new GlobalMaxPool1DLayer();
pool.SetInput(data);
}
public void Test_Cropping1D_Not1DInput_Trim()
{
Data2D data = new Data2D(2, 4, 3, 5);
Cropping1DLayer crop = new Cropping1DLayer(1, 1);
crop.SetInput(data);
}
public void Test_GRU_Null_Weights()
{
Data2D weights = null;
GRULayer rnn = new GRULayer(5, 3, TanHLayer.TanHLambda, TanHLayer.TanHLambda);
rnn.SetWeights(weights);
}
public void Test_GRU_WrongSizeinWidth_Weights()
{
Data2D weights = new Data2D(1, 3, 5, 4);
GRULayer rnn = new GRULayer(5, 3, TanHLayer.TanHLambda, p => { });
rnn.SetWeights(weights);
}
public void Test_Sigmoid_KerasModel()
{
string path = @"tests\test_sigmoid_model.json";
var reader = new ReaderKerasModel(path);
SequentialModel model = reader.GetSequentialExecutor();
Data2D inp = new Data2D(1, 8, 1, 1);
inp[0, 0, 0, 0] = 1;
inp[0, 1, 0, 0] = 2;
inp[0, 2, 0, 0] = -1;
inp[0, 3, 0, 0] = 0;
inp[0, 4, 0, 0] = 3;
inp[0, 5, 0, 0] = 1;
inp[0, 6, 0, 0] = 1;
inp[0, 7, 0, 0] = 2;
Data2D ou = model.ExecuteNetwork(inp) as Data2D;
Assert.AreEqual(ou.GetDimension().c, 4);
Assert.AreEqual(ou.GetDimension().w, 1);
Assert.AreEqual(ou[0, 0, 0, 0], 0.2689414322376251, 0.00001);
Assert.AreEqual(ou[0, 0, 1, 0], 0.9959298968315125, 0.00001);
Assert.AreEqual(ou[0, 0, 2, 0], 1.0, 0.00001);
Assert.AreEqual(ou[0, 0, 3, 0], 0.9998766183853149, 0.00001);
}
public void Test_Softmax_KerasModel()
{
string path = @"tests\test_softmax_model.json";
var reader = new ReaderKerasModel(path);
SequentialModel model = reader.GetSequentialExecutor();
Data2D inp = new Data2D(1, 8, 1, 1);
inp[0, 0, 0, 0] = 1;
inp[0, 1, 0, 0] = 2;
inp[0, 2, 0, 0] = -1;
inp[0, 3, 0, 0] = 0;
inp[0, 4, 0, 0] = 3;
inp[0, 5, 0, 0] = 1;
inp[0, 6, 0, 0] = 1;
inp[0, 7, 0, 0] = 2;
Data2D ou = model.ExecuteNetwork(inp) as Data2D;
Assert.AreEqual(ou.GetDimension().c, 4);
Assert.AreEqual(ou.GetDimension().w, 1);
Assert.AreEqual(ou[0, 0, 0, 0], 3.3980058766758248e-09, 1e-10);
Assert.AreEqual(ou[0, 0, 1, 0], 2.26015504267707e-06, 1e-7);
Assert.AreEqual(ou[0, 0, 2, 0], 0.9999228715896606, 0.00001);
Assert.AreEqual(ou[0, 0, 3, 0], 7.484605885110795e-05, 1e-6);
}
public void Test_Softmax_DifferentData()
{
Data2D data = new Data2D(5, 4, 5, 10);
SoftmaxLayer soft = new SoftmaxLayer();
soft.SetInput(data);
}
public void Test_Softsign_Execute()
{
softsign = new SoftsignLayer();
Data2D data = new Data2D(2, 3, 1, 1);
data[0, 0, 0, 0] = 4;
data[0, 1, 0, 0] = 2;
data[0, 2, 0, 0] = -2;
data[1, 0, 0, 0] = 3;
data[1, 1, 0, 0] = -1;
data[1, 2, 0, 0] = -3;
softsign.SetInput(data);
softsign.Execute();
Data2D output = softsign.GetOutput() as Data2D;
Assert.AreEqual(output[0, 0, 0, 0], SoftsignFunc(4.0), 0.00000001);
Assert.AreEqual(output[0, 1, 0, 0], SoftsignFunc(2.0), 0.00000001);
Assert.AreEqual(output[0, 2, 0, 0], SoftsignFunc(-2.0), 0.00000001);
Assert.AreEqual(output[1, 0, 0, 0], SoftsignFunc(3.0), 0.00000001);
Assert.AreEqual(output[1, 1, 0, 0], SoftsignFunc(-1.0), 0.00000001);
Assert.AreEqual(output[1, 2, 0, 0], SoftsignFunc(-3.0), 0.00000001);
}
public void Test_AvgPool1D_1_KerasModel()
{
string path = @"tests\test_avgpool_1D_1_model.json";
var reader = new ReaderKerasModel(path);
SequentialModel model = reader.GetSequentialExecutor();
Data2D inp = new Data2D(1, 5, 2, 1);
inp[0, 0, 0, 0] = 0;
inp[0, 0, 1, 0] = 1;
inp[0, 1, 0, 0] = 2;
inp[0, 1, 1, 0] = 1;
inp[0, 2, 0, 0] = 0;
inp[0, 2, 1, 0] = 0;
inp[0, 3, 0, 0] = 2;
inp[0, 3, 1, 0] = 1;
inp[0, 4, 0, 0] = 2;
inp[0, 4, 1, 0] = 1;
Data2D ou = model.ExecuteNetwork(inp) as Data2D;
Assert.AreEqual(ou.GetDimension().c, 2);
Assert.AreEqual(ou.GetDimension().w, 3);
Assert.AreEqual(ou[0, 0, 0, 0], 0.6666666865348816, 0.00001);
Assert.AreEqual(ou[0, 0, 1, 0], 0.6666666865348816, 0.00001);
Assert.AreEqual(ou[0, 1, 0, 0], 1.3333333730697632, 0.00001);
Assert.AreEqual(ou[0, 1, 1, 0], 0.6666666865348816, 0.00001);
Assert.AreEqual(ou[0, 2, 0, 0], 1.3333333730697632, 0.00001);
Assert.AreEqual(ou[0, 2, 1, 0], 0.6666666865348816, 0.00001);
}
public void Test_Cropping2D_Null_Input()
{
Data2D data = null;
Cropping2DLayer crop = new Cropping2DLayer(1, 2, 1, 1);
crop.SetInput(data);
}
public void Test_RepeatVector_WrongSizesWidth()
{
Data2D data = new Data2D(1, 3, 5, 2);
RepeatVectorLayer rep = new RepeatVectorLayer(3);
rep.SetInput(data);
}
public void Test_Cropping2D_TooMuchCropping_Trim()
{
Data2D data = new Data2D(4, 4, 3, 5);
Cropping2DLayer crop = new Cropping2DLayer(2, 3, 1, 1);
crop.SetInput(data);
}
public void Test_RepeatVector_Null_Input()
{
Data2D data = null;
RepeatVectorLayer rep = new RepeatVectorLayer(4);
rep.SetInput(data);
}
public void Test_HardSigmoid_Execute()
{
hardsigmoid = new HardSigmoidLayer();
Data2D data = new Data2D(2, 3, 1, 1);
data[0, 0, 0, 0] = 4;
data[0, 1, 0, 0] = 2;
data[0, 2, 0, 0] = -2;
data[1, 0, 0, 0] = 3;
data[1, 1, 0, 0] = -1;
data[1, 2, 0, 0] = -3;
hardsigmoid.SetInput(data);
hardsigmoid.Execute();
Data2D output = hardsigmoid.GetOutput() as Data2D;
Assert.AreEqual(output[0, 0, 0, 0], HardSigmoidFunc(4.0), 0.00000001);
Assert.AreEqual(output[0, 1, 0, 0], HardSigmoidFunc(2.0), 0.00000001);
Assert.AreEqual(output[0, 2, 0, 0], HardSigmoidFunc(-2.0), 0.00000001);
Assert.AreEqual(output[1, 0, 0, 0], HardSigmoidFunc(3.0), 0.00000001);
Assert.AreEqual(output[1, 1, 0, 0], HardSigmoidFunc(-1.0), 0.00000001);
Assert.AreEqual(output[1, 2, 0, 0], HardSigmoidFunc(-3.0), 0.00000001);
}
public void Test_RepeatVector_WrongSizesHeight()
{
Data2D data = new Data2D(2, 1, 5, 2);
RepeatVectorLayer rep = new RepeatVectorLayer(3);
rep.SetInput(data);
}
public void Test_MaxPool1D_Null_Input()
{
Data2D data = null;
MaxPool1DLayer pool = new MaxPool1DLayer(1, 1, 2);
pool.SetInput(data);
}
public void Test_Conv1D_NullConv_Weights()
{
Data2D weights = null;
Conv1DLayer conv = new Conv1DLayer(1, 1);
conv.SetWeights(weights);
}
public void Test_ReLu_KerasModel()
{
string path = @"tests\test_relu_model.json";
var reader = new ReaderKerasModel(path);
SequentialModel model = reader.GetSequentialExecutor();
Data2D inp = new Data2D(1, 8, 1, 1);
inp[0, 0, 0, 0] = 1;
inp[0, 1, 0, 0] = 2;
inp[0, 2, 0, 0] = -1;
inp[0, 3, 0, 0] = 0;
inp[0, 4, 0, 0] = 3;
inp[0, 5, 0, 0] = 1;
inp[0, 6, 0, 0] = 1;
inp[0, 7, 0, 0] = 2;
Data2D ou = model.ExecuteNetwork(inp) as Data2D;
Assert.AreEqual(ou.GetDimension().c, 4);
Assert.AreEqual(ou.GetDimension().w, 1);
Assert.AreEqual(ou[0, 0, 0, 0], 0.0, 0.00001);
Assert.AreEqual(ou[0, 0, 1, 0], 5.5, 0.00001);
Assert.AreEqual(ou[0, 0, 2, 0], 18.5, 0.00001);
Assert.AreEqual(ou[0, 0, 3, 0], 9.0, 0.00001);
}
public void SetInput(IData input)
{
if (input == null)
{
throw new Exception("BatchNormLayer: input is null.");
}
else if (!(input is Data2D))
{
throw new Exception("BatchNormLayer: input is not Data2D.");
}
this.input = input as Data2D;
Dimension dimI = this.input.GetDimension();
int kGamma = this.gamma.Count;
int kBeta = this.beta.Count;
int kBias = this.bias.Count;
int kVariance = this.variance.Count;
if (dimI.c != kBias || dimI.c != kGamma || dimI.c != kBeta || dimI.c != kVariance)
{
throw new Exception("Number of parameters is not equal with number of features (channels).");
}
int outputH = dimI.h;
int outputW = dimI.w;
int outputC = dimI.c;
int outputB = dimI.b;
output = new Data2D(outputH, outputW, outputC, outputB);
}
public void Test_Reshape2D_Null_Input()
{
Data2D data = null;
Reshape2DLayer res = new Reshape2DLayer(1, 2, 1, 4);
res.SetInput(data);
}
public void SetWeights(IData parameters)
{
if (parameters == null)
{
throw new Exception("BatchNormLayer: parameters is null.");
}
else if (!(parameters is Data2D))
{
throw new Exception("BatchNormLayer: parameters is not Data2D.");
}
Data2D pms = parameters as Data2D;
if (pms.GetDimension().h != 1 || pms.GetDimension().w != 1)
{
throw new Exception("BatchNormLayer: parameters' height and width should be 1.");
}
for (int feature = 0; feature < pms.GetDimension().c; ++feature)
{
gamma.Add(pms[0, 0, feature, 0]);
beta.Add(pms[0, 0, feature, 1]);
bias.Add(pms[0, 0, feature, 2]);
variance.Add(pms[0, 0, feature, 3]);
}
}
public static void Main(string[] args)
{
// Keras speed with the same: 60 ms.
/*ReaderKerasModel reader = new ReaderKerasModel("test_cnn_model.json");
* SequentialModel model = reader.GetSequentialExecutor();
*
* Console.WriteLine((model.GetSummary() as SequentialModelData).GetStringRepresentation());
*
* Console.ReadKey();
* int[] idx = { 1,2,3};
*
* Console.WriteLine(idx[1]);
* Console.ReadKey();*/
Conv2DLayer layer = new Conv2DLayer(0, 0, 1, 1);
Data2D input = new Data2D(6, 5, 3, 1);
input[0, 0, 0, 0] = 1; input[0, 1, 0, 0] = 2; input[0, 2, 0, 0] = 2; input[0, 3, 0, 0] = 1; input[0, 4, 0, 0] = 4;
input[1, 0, 0, 0] = 3; input[1, 1, 0, 0] = 1; input[1, 2, 0, 0] = 0; input[1, 3, 0, 0] = 2; input[1, 4, 0, 0] = 1;
input[2, 0, 0, 0] = 0; input[2, 1, 0, 0] = 2; input[2, 2, 0, 0] = 2; input[2, 3, 0, 0] = 5; input[2, 4, 0, 0] = 2;
input[3, 0, 0, 0] = 6; input[3, 1, 0, 0] = -2; input[3, 2, 0, 0] = -1; input[3, 3, 0, 0] = 3; input[3, 4, 0, 0] = 1;
input[4, 0, 0, 0] = 2; input[4, 1, 0, 0] = 1; input[4, 2, 0, 0] = 2; input[4, 3, 0, 0] = 4; input[4, 4, 0, 0] = 0;
input[5, 0, 0, 0] = 5; input[5, 1, 0, 0] = -3; input[5, 2, 0, 0] = -1; input[5, 3, 0, 0] = -4; input[5, 4, 0, 0] = 0;
input[0, 0, 1, 0] = 2; input[0, 1, 1, 0] = 0; input[0, 2, 1, 0] = 2; input[0, 3, 1, 0] = -1; input[0, 4, 1, 0] = 3;
input[1, 0, 1, 0] = 2; input[1, 1, 1, 0] = 5; input[1, 2, 1, 0] = -1; input[1, 3, 1, 0] = 3; input[1, 4, 1, 0] = 5;
input[2, 0, 1, 0] = 1; input[2, 1, 1, 0] = 1; input[2, 2, 1, 0] = 1; input[2, 3, 1, 0] = 0; input[2, 4, 1, 0] = 1;
input[3, 0, 1, 0] = -3; input[3, 1, 1, 0] = 2; input[3, 2, 1, 0] = -1; input[3, 3, 1, 0] = 4; input[3, 4, 1, 0] = 1;
input[4, 0, 1, 0] = 2; input[4, 1, 1, 0] = 1; input[4, 2, 1, 0] = 2; input[4, 3, 1, 0] = 2; input[4, 4, 1, 0] = 1;
input[5, 0, 1, 0] = 0; input[5, 1, 1, 0] = -3; input[5, 2, 1, 0] = 1; input[5, 3, 1, 0] = -2; input[5, 4, 1, 0] = -1;
input[0, 0, 2, 0] = 4; input[0, 1, 2, 0] = 5; input[0, 2, 2, 0] = 0; input[0, 3, 2, 0] = -1; input[0, 4, 2, 0] = -3;
input[1, 0, 2, 0] = 2; input[1, 1, 2, 0] = 3; input[1, 2, 2, 0] = 1; input[1, 3, 2, 0] = 6; input[1, 4, 2, 0] = 0;
input[2, 0, 2, 0] = 0; input[2, 1, 2, 0] = -4; input[2, 2, 2, 0] = -3; input[2, 3, 2, 0] = -2; input[2, 4, 2, 0] = -4;
input[3, 0, 2, 0] = 4; input[3, 1, 2, 0] = 2; input[3, 2, 2, 0] = 1; input[3, 3, 2, 0] = 0; input[3, 4, 2, 0] = 4;
input[4, 0, 2, 0] = 3; input[4, 1, 2, 0] = 3; input[4, 2, 2, 0] = 0; input[4, 3, 2, 0] = 1; input[4, 4, 2, 0] = 1;
input[5, 0, 2, 0] = -2; input[5, 1, 2, 0] = 1; input[5, 2, 2, 0] = 1; input[5, 3, 2, 0] = 0; input[5, 4, 2, 0] = 5;
Data2D kernel = new Data2D(3, 3, 3, 1);
kernel[0, 0, 0, 0] = 1; kernel[0, 1, 0, 0] = 1; kernel[0, 2, 0, 0] = 0;
kernel[1, 0, 0, 0] = 2; kernel[1, 1, 0, 0] = 0; kernel[1, 2, 0, 0] = 0;
kernel[2, 0, 0, 0] = 1; kernel[2, 1, 0, 0] = 2; kernel[2, 2, 0, 0] = 1;
kernel[0, 0, 1, 0] = 3; kernel[0, 1, 1, 0] = 1; kernel[0, 2, 1, 0] = -1;
kernel[1, 0, 1, 0] = 2; kernel[1, 1, 1, 0] = -1; kernel[1, 2, 1, 0] = -2;
kernel[2, 0, 1, 0] = 0; kernel[2, 1, 1, 0] = 1; kernel[2, 2, 1, 0] = 2;
kernel[0, 0, 2, 0] = 0; kernel[0, 1, 2, 0] = 1; kernel[0, 2, 2, 0] = 1;
kernel[1, 0, 2, 0] = -1; kernel[1, 1, 2, 0] = 2; kernel[1, 2, 2, 0] = 1;
kernel[2, 0, 2, 0] = 3; kernel[2, 1, 2, 0] = 0; kernel[2, 2, 2, 0] = 1;
layer.SetWeights(kernel);
layer.SetInput(input);
layer.Execute();
Data2D output = layer.GetOutput() as Data2D;
}
public void Test_LSTM_WrongSizeinChannel_Weights()
{
Data2D weights = new Data2D(1, 5, 7, 4);
LSTMLayer rnn = new LSTMLayer(5, 3, TanHLayer.TanHLambda, p => { });
rnn.SetWeights(weights);
}