private void AssertArgumentException(IEvaluateModelManagedF model,
Bitmap image,
string outputKey,
string expectedParameterName,
string expectedMessageText,
string errorMessage)
{
bool exception = false;
try
{
model.EvaluateRgbImage(image, outputKey);
}
catch (ArgumentException ex)
{
if (ex.ParamName == expectedParameterName && ex.Message.Contains(expectedMessageText))
{
exception = true;
}
}
catch { }
if (!exception)
{
throw new Exception(errorMessage);
}
}
public void EvalManagedImageApiErrorHandling()
{
// The width and height of the image that will be fed into the network.
var expectedSize = 10;
// Images with correct size and pixel format.
var correctBmp1 = new Bitmap(expectedSize, expectedSize, PixelFormat.Format24bppRgb);
var correctBmp2 = new Bitmap(expectedSize, expectedSize, PixelFormat.Format32bppRgb);
// Image with correct size, but wrong pixel format
var wrongPixelFormat = new Bitmap(expectedSize, expectedSize, PixelFormat.Format16bppRgb565);
// Image with wrong size, correct pixel format
var wrongSize = new Bitmap(expectedSize * 2, expectedSize, PixelFormat.Format24bppRgb);
var inputVectorSize = expectedSize * expectedSize * 3;
var modelDefinition = String.Format(@"deviceId = -1
precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
i1 = Input({0}) # Network must have size expectedSize * expectedSize * 3, for 3 channels
o1 = Times(Constant(1, rows=2, cols={0}), i1, tag=""output"")
FeatureNodes = (i1)
]", inputVectorSize);
using (var model = new IEvaluateModelManagedF())
{
model.CreateNetwork(modelDefinition);
var output = model.EvaluateRgbImage(correctBmp1, "o1");
// Network computes 2 simple dot products.
Assert.AreEqual(2, output.Count, "Size of output vector");
// Input image is all zero, output should be too.
Assert.AreEqual(0.0f, output[0], "OutputVector[0]");
Assert.AreEqual(0.0f, output[1], "OutputVector[1]");
AssertArgumentException(model,
correctBmp1,
"No such output key",
"outputKey",
"not an output node",
"Providing a non-existing output node should fail with an ArgumentException.");
AssertArgumentException(model,
wrongPixelFormat,
"o1",
"image",
"must be one of { Format24bppRgb, Format32bppArgb}",
"Images with an unrecognized pixel format should fail with an ArgumentException.");
AssertArgumentException(model,
wrongSize,
"o1",
"image",
"invalid size",
"Calling with a wrongly sized image should fail with an ArgumentException.");
}
}
/// <summary>
/// This method shows how to evaluate a trained image classification model, where the
/// creation of the CNTK feature vector is happening in native code inside the EvalWrapper.
/// </summary>
public static List <float> EvaluateImageInputUsingImageApi()
{
List <float> outputs = null;
try
{
// This example requires the RestNet_18 model.
// The model can be downloaded from <see cref="https://www.cntk.ai/resnet/ResNet_18.model"/>
// The model is assumed to be located at: <CNTK>\Examples\Image\Classification\ResNet
// along with a sample image file named "zebra.jpg".
Environment.CurrentDirectory = initialDirectory;
using (var model = new IEvaluateModelManagedF())
{
model.CreateNetwork(string.Format("modelPath=\"{0}\"", resnetModelFilePath), deviceId: -1);
// Prepare input value in the appropriate structure and size
var inDims = model.GetNodeDimensions(NodeGroup.Input);
if (inDims.First().Value != resNetImageSize * resNetImageSize * 3)
{
throw new CNTKRuntimeException(string.Format("The input dimension for {0} is {1} which is not the expected size of {2}.", inDims.First(), inDims.First().Value, 224 * 224 * 3), string.Empty);
}
// Transform the image
Bitmap bmp = new Bitmap(Bitmap.FromFile(imageFileName));
var resized = bmp.Resize(resNetImageSize, resNetImageSize, true);
// Now evaluate using the alternative API, where we directly pass the
// native bitmap data to the unmanaged code.
var outDims = model.GetNodeDimensions(NodeGroup.Output);
var outputNodeName = outDims.First().Key;
outputs = model.EvaluateRgbImage(resized, outputNodeName);
}
// Retrieve the outcome index (so we can compare it with the expected index)
var max = outputs.Select((value, index) => new { Value = value, Index = index })
.Aggregate((a, b) => (a.Value > b.Value) ? a : b)
.Index;
Console.WriteLine("EvaluateImageInputUsingImageApi: Outcome = {0}", max);
}
catch (CNTKException ex)
{
OnCNTKException(ex);
}
catch (Exception ex)
{
OnGeneralException(ex);
}
return(outputs);
}
/// <summary>
/// This method shows how to evaluate a trained image classification model, where the
/// creation of the CNTK feature vector is happening in native code inside the EvalWrapper.
/// </summary>
public static List<float> EvaluateImageInputUsingImageApi()
{
List<float> outputs = null;
try
{
// This example requires the RestNet_18 model.
// The model can be downloaded from <see cref="https://www.cntk.ai/resnet/ResNet_18.model"/>
// The model is assumed to be located at: <CNTK>\Examples\Image\Classification\ResNet
// along with a sample image file named "zebra.jpg".
Environment.CurrentDirectory = initialDirectory;
using (var model = new IEvaluateModelManagedF())
{
model.CreateNetwork(string.Format("modelPath=\"{0}\"", resnetModelFilePath), deviceId: -1);
// Prepare input value in the appropriate structure and size
var inDims = model.GetNodeDimensions(NodeGroup.Input);
if (inDims.First().Value != resNetImageSize * resNetImageSize * 3)
{
throw new CNTKRuntimeException(string.Format("The input dimension for {0} is {1} which is not the expected size of {2}.", inDims.First(), inDims.First().Value, 224 * 224 * 3), string.Empty);
}
// Transform the image
Bitmap bmp = new Bitmap(Bitmap.FromFile(imageFileName));
var resized = bmp.Resize(resNetImageSize, resNetImageSize, true);
// Now evaluate using the alternative API, where we directly pass the
// native bitmap data to the unmanaged code.
var outDims = model.GetNodeDimensions(NodeGroup.Output);
var outputNodeName = outDims.First().Key;
outputs = model.EvaluateRgbImage(resized, outputNodeName);
}
// Retrieve the outcome index (so we can compare it with the expected index)
var max = outputs.Select((value, index) => new { Value = value, Index = index })
.Aggregate((a, b) => (a.Value > b.Value) ? a : b)
.Index;
Console.WriteLine("EvaluateImageInputUsingImageApi: Outcome = {0}", max);
}
catch (CNTKException ex)
{
OnCNTKException(ex);
}
catch (Exception ex)
{
OnGeneralException(ex);
}
return outputs;
}
