public OnnxModelConfigurator(IOnnxModel onnxModel)
{
mlContext = new MLContext();
// Model creation and pipeline definition for images needs to run just once,
// so calling it from the constructor:
mlModel = SetupMlNetModel(onnxModel);
}
private ITransformer SetupMlNetModel(IOnnxModel onnxModel)
{
var dataView = _mlContext.Data
.LoadFromEnumerable(new List <TFeature>());
var pipeline = _mlContext.Transforms
.ApplyOnnxModel(modelFile: onnxModel.ModelPath, outputColumnNames: onnxModel.ModelOutput, inputColumnNames: onnxModel.ModelInput, gpuDeviceId: 0);
var mlNetModel = pipeline.Fit(dataView);
return(mlNetModel);
}
private ITransformer SetupMlNetModel(IOnnxModel onnxModel)
{
var dataView = mlContext.Data.LoadFromEnumerable(new List <ImageInputData>());
var pipeline = mlContext.Transforms.ResizeImages(resizing: ImageResizingEstimator.ResizingKind.Fill, outputColumnName: onnxModel.ModelInput, imageWidth: ImageSettings.imageWidth, imageHeight: ImageSettings.imageHeight, inputColumnName: nameof(ImageInputData.Image))
.Append(mlContext.Transforms.ExtractPixels(outputColumnName: onnxModel.ModelInput))
.Append(mlContext.Transforms.ApplyOnnxModel(modelFile: onnxModel.ModelPath, outputColumnName: onnxModel.ModelOutput, inputColumnName: onnxModel.ModelInput));
var mlNetModel = pipeline.Fit(dataView);
return(mlNetModel);
}
private ITransformer SetupMlNetModel(IOnnxModel onnxModel)
{
var dataView = mlContext.Data.LoadFromEnumerable(new List <ImageInputData>());
// If the input flattened image data order is different from im.read() check "interleavePixelColors" to true
var pipeline = mlContext.Transforms.ResizeImages(resizing: ImageResizingEstimator.ResizingKind.Fill, outputColumnName: onnxModel.ModelInput, imageWidth: ImageSettings.imageWidth, imageHeight: ImageSettings.imageHeight, inputColumnName: nameof(ImageInputData.Image))
.Append(mlContext.Transforms.ExtractPixels(outputColumnName: onnxModel.ModelInput, interleavePixelColors: true, orderOfExtraction: ImagePixelExtractingEstimator.ColorsOrder.ARGB, scaleImage: 1.0f / 255.0f))
.Append(mlContext.Transforms.ApplyOnnxModel(modelFile: onnxModel.ModelPath, outputColumnName: onnxModel.ModelOutput, inputColumnName: onnxModel.ModelInput));
var mlNetModel = pipeline.Fit(dataView);
return(mlNetModel);
}
public ITransformer LoadOnnx(IOnnxModel model)
{
var dataview = ml.Data.LoadFromEnumerable(new List <ImageNetData>());
var estimator = ml.Transforms.ResizeImages(
inputColumnName: nameof(ImageNetData.Image),
outputColumnName: model.InputName,
imageWidth: ImageNetSettings.Width,
imageHeight: ImageNetSettings.Height,
resizing: ResizingKind.Fill
)
.Append(ml.Transforms.ExtractPixels(outputColumnName: model.InputName))
.Append(ml.Transforms.ApplyOnnxModel(model.OutputName, model.InputName, model.ModelPath, 1, true));
var transformer = estimator.Fit(dataview);
return(transformer);
}
public OnnxModelConfigurator(IOnnxModel onnxModel)
{
_mlContext = new MLContext();
_mlModel = SetupMlNetModel(onnxModel);
}