CNTK.MinibatchSource create_minibatch_source(CNTK.NDShape shape, int start_index, int sample_count, string prefix, bool is_training = false, bool use_augmentations = false)
{
var map_filename = create_map_txt_file_if_needed(start_index, sample_count, prefix);
var transforms = new List <CNTK.CNTKDictionary>();
if (use_augmentations)
{
var randomSideTransform = CNTK.CNTKLib.ReaderCrop("RandomSide",
new Tuple <int, int>(0, 0),
new Tuple <float, float>(0.8f, 1.0f),
new Tuple <float, float>(0.0f, 0.0f),
new Tuple <float, float>(1.0f, 1.0f),
"uniRatio");
transforms.Add(randomSideTransform);
}
var scaleTransform = CNTK.CNTKLib.ReaderScale(width: shape[1], height: shape[0], channels: shape[2]);
transforms.Add(scaleTransform);
var imageDeserializer = CNTK.CNTKLib.ImageDeserializer(map_filename, "labels", 2, "features", transforms);
var minibatchSourceConfig = new CNTK.MinibatchSourceConfig(new CNTK.DictionaryVector()
{
imageDeserializer
});
if (!is_training)
{
minibatchSourceConfig.randomizationWindowInChunks = 0;
minibatchSourceConfig.randomizationWindowInSamples = 0;
}
return(CNTK.CNTKLib.CreateCompositeMinibatchSource(minibatchSourceConfig));
}
public static CNTK.Value get_tensors(CNTK.NDShape shape, float[] src, int[] indices, int indices_begin, int indices_end, CNTK.DeviceDescriptor device)
{
var cpu_tensors = Util.get_minibatch_data_CPU(shape, src, indices, indices_begin, indices_end);
var result = CNTK.Value.CreateBatch(shape, cpu_tensors, device, true);
return(result);
}
/// <summary>
/// Add a pooling layer to a neural network.
/// </summary>
/// <param name="input">The neural network to expand</param>
/// <param name="poolingType">The type of pooling to perform</param>
/// <param name="windowShape">The shape of the pooling window</param>
/// <param name="strides">The stride lengths</param>
/// <returns>The neural network with the pooling layer added.</returns>
public static CNTK.Variable Pooling(
this CNTK.Variable input,
CNTK.PoolingType poolingType,
CNTK.NDShape windowShape,
int[] strides)
{
return(CNTK.CNTKLib.Pooling(input, poolingType, windowShape, strides));
}
static float[] get_minibatch_data_CPU(CNTK.NDShape shape, float[] src, int indices_begin, int indices_end)
{
// it would be nice if we avoid the copy here
var result = new float[indices_end - indices_begin];
Array.Copy(src, indices_begin, result, 0, result.Length);
return(result);
}
static float[] get_minibatch_data_CPU(CNTK.NDShape shape, float[] src, int[] indices, int indices_begin, int indices_end)
{
var num_indices = indices_end - indices_begin;
var row_length = shape.TotalSize;
var result = new float[num_indices];
var row_index = 0;
for (var index = indices_begin; index != indices_end; index++)
{
result[row_index++] = src[indices[index]];
}
return(result);
}
static CNTK.NDArrayView[] get_minibatch_data_CPU(CNTK.NDShape shape, float[][] src, int indices_begin, int indices_end)
{
var num_indices = indices_end - indices_begin;
var result = new CNTK.NDArrayView[num_indices];
var row_index = 0;
for (var index = indices_begin; index != indices_end; index++)
{
var dataBuffer = src[index];
var ndArrayView = new CNTK.NDArrayView(shape, dataBuffer, CNTK.DeviceDescriptor.CPUDevice, true);
result[row_index++] = ndArrayView;
}
return(result);
}
/// <summary>
/// Reshape the current network tensor to the new shape.
/// </summary>
/// <param name="input">The neural network</param>
/// <param name="newShape">The new shape to reshape the tensor to</param>
/// <returns>The neural network with the reshape layer added</returns>
public static CNTK.Variable Reshape(
this CNTK.Variable input,
CNTK.NDShape newShape)
{
return(CNTK.CNTKLib.Reshape(input, newShape));
}
