public void BuildAndTrain()
{
var featurizerModelLocation = inputModelLocation;
ConsoleHelpers.ConsoleWriteHeader("Read model");
Console.WriteLine($"Model location: {featurizerModelLocation}");
Console.WriteLine($"Images folder: {imagesFolder}");
Console.WriteLine($"Training file: {dataLocation}");
Console.WriteLine($"Default parameters: image size=({ImageNetSettings.imageWidth},{ImageNetSettings.imageHeight}), image mean: {ImageNetSettings.mean}");
var data = mlContext.Data.LoadFromTextFile <ImageNetData>(path: dataLocation, hasHeader: false);
var pipeline = mlContext.Transforms.Conversion.MapValueToKey(outputColumnName: LabelTokey, inputColumnName: "Label")
.Append(mlContext.Transforms.LoadImages(outputColumnName: "input", imageFolder: imagesFolder, inputColumnName: nameof(ImageNetData.ImagePath)))
.Append(mlContext.Transforms.ResizeImages(outputColumnName: "input", imageWidth: ImageNetSettings.imageWidth, imageHeight: ImageNetSettings.imageHeight, inputColumnName: "input"))
.Append(mlContext.Transforms.ExtractPixels(outputColumnName: "input", interleavePixelColors: ImageNetSettings.channelsLast, offsetImage: ImageNetSettings.mean))
.Append(mlContext.Model.LoadTensorFlowModel(featurizerModelLocation).
ScoreTensorFlowModel(outputColumnNames: new[] { "softmax2_pre_activation" }, inputColumnNames: new[] { "input" }, addBatchDimensionInput: true))
.Append(mlContext.MulticlassClassification.Trainers.LbfgsMaximumEntropy(labelColumnName: LabelTokey, featureColumnName: "softmax2_pre_activation"))
.Append(mlContext.Transforms.Conversion.MapKeyToValue(PredictedLabelValue, "PredictedLabel"))
.AppendCacheCheckpoint(mlContext);
// Train the model
ConsoleHelpers.ConsoleWriteHeader("Training classification model");
ITransformer model = pipeline.Fit(data);
// Process the training data through the model
// This is an optional step, but it's useful for debugging issues
var trainData = model.Transform(data);
var loadedModelOutputColumnNames = trainData.Schema
.Where(col => !col.IsHidden).Select(col => col.Name);
var trainData2 = mlContext.Data.CreateEnumerable <ImageNetPipeline>(trainData, false, true).ToList();
trainData2.ForEach(pr => ConsoleHelpers.ConsoleWriteImagePrediction(pr.ImagePath, pr.PredictedLabelValue, pr.Score.Max()));
// Get some performance metric on the model using training data
var classificationContext = mlContext.MulticlassClassification;
ConsoleHelpers.ConsoleWriteHeader("Classification metrics");
var metrics = classificationContext.Evaluate(trainData, labelColumnName: LabelTokey, predictedLabelColumnName: "PredictedLabel");
Console.WriteLine($"LogLoss is: {metrics.LogLoss}");
Console.WriteLine($"PerClassLogLoss is: {String.Join(" , ", metrics.PerClassLogLoss.Select(c => c.ToString()))}");
// Save the model to assets/outputs
ConsoleHelpers.ConsoleWriteHeader("Save model to local file");
ModelHelpers.DeleteAssets(outputModelLocation);
mlContext.Model.Save(model, trainData.Schema, outputModelLocation);
Console.WriteLine($"Model saved: {outputModelLocation}");
}
public void ClassifyImages()
{
ConsoleHelpers.ConsoleWriteHeader("Loading model");
Console.WriteLine($"Model loaded: {modelLocation}");
// Load the model
ITransformer loadedModel = mlContext.Model.Load(modelLocation, out var modelInputSchema);
// Make prediction function (input = ImageNetData, output = ImageNetPrediction)
var predictor = mlContext.Model.CreatePredictionEngine <ImageNetData, ImageNetPrediction>(loadedModel);
// Read csv file into List<ImageNetData>
var imageListToPredict = ImageNetData.ReadFromCsv(dataLocation, imagesFolder).ToList();
ConsoleHelpers.ConsoleWriteHeader("Making classifications");
// There is a bug (https://github.com/dotnet/machinelearning/issues/1138),
// that always buffers the response from the predictor
// so we have to make a copy-by-value op everytime we get a response
// from the predictor
imageListToPredict
.Select(td => new { td, pred = predictor.Predict(td) })
.Select(pr => (pr.td.ImagePath, pr.pred.PredictedLabelValue, pr.pred.Score))
.ToList()
.ForEach(pr => ConsoleHelpers.ConsoleWriteImagePrediction(pr.ImagePath, pr.PredictedLabelValue, pr.Score.Max()));
}
