title | description | author | ms.author | ms.date | ms.service | ms.subservice | ms.topic | keywords | monikerRange | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
featurizeImage function (MicrosoftML) |
Featurizes an image using a pre-trained deep neural network model (MicrosoftML). |
rothja |
jroth |
07/15/2019 |
sql |
machine-learning |
reference |
|
>=sql-server-2016||>=sql-server-linux-ver15 |
Featurizes an image using a pre-trained deep neural network model.
featurizeImage(var, outVar = NULL, dnnModel = "Resnet18")
Input variable containing extracted pixel values.
The prefix of the output variables containing the image features. If null, the input variable name will be used. The default value is NULL
.
The pre-trained deep neural network. The possible options are:
"resnet18"
"resnet50"
"resnet101"
"alexnet"
The default value is"resnet18"
. SeeDeep Residual Learning for Image Recognition
for details about ResNet.
featurizeImage
featurizes an image using the specified
pre-trained deep neural network model. The input variables to this transform must be extracted pixel values.
A maml
object defining the transform.
Microsoft Corporation Microsoft Technical Support
train <- data.frame(Path = c(system.file("help/figures/RevolutionAnalyticslogo.png", package = "MicrosoftML")), Label = c(TRUE), stringsAsFactors = FALSE)
# Loads the images from variable Path, resizes the images to 1x1 pixels and trains a neural net.
model <- rxNeuralNet(
Label ~ Features,
data = train,
mlTransforms = list(
loadImage(vars = list(Features = "Path")),
resizeImage(vars = "Features", width = 1, height = 1, resizing = "Aniso"),
extractPixels(vars = "Features")
),
mlTransformVars = "Path",
numHiddenNodes = 1,
numIterations = 1)
# Featurizes the images from variable Path using the default model, and trains a linear model on the result.
model <- rxFastLinear(
Label ~ Features,
data = train,
mlTransforms = list(
loadImage(vars = list(Features = "Path")),
resizeImage(vars = "Features", width = 224, height = 224), # If dnnModel == "AlexNet", the image has to be resized to 227x227.
extractPixels(vars = "Features"),
featurizeImage(var = "Features")
),
mlTransformVars = "Path")