DeepLearning_ImageClassification_TensorFlow

Image Classification - Scoring sample

ML.NET version	API type	Status	App Type	Data type	Scenario	ML Task	Algorithms
v1.5.0	Dynamic API	up-to-date	Console app	Images and text labels	Images classification	TensorFlow Inception5h	DeepLearning model

Problem

Image classification is a common case in many business scenarios. For these cases, you can either use pre-trained models or train your own model to classify images specific to your custom domain.

DataSet

There are two data sources: the tsv file and the image files. The tsv file contains two columns: the first one is defined as ImagePath and the second one is the Label corresponding to the image. As you can observe, the file does not have a header row, and looks like this:

broccoli.jpg	broccoli
broccoli.png	broccoli
canoe2.jpg	canoe
canoe3.jpg	canoe
canoe4.jpg	canoe
coffeepot.jpg	coffeepot
coffeepot2.jpg	coffeepot
coffeepot3.jpg	coffeepot
coffeepot4.jpg	coffeepot
pizza.jpg	pizza
pizza2.jpg	pizza
pizza3.jpg	pizza
teddy1.jpg	teddy bear
teddy2.jpg	teddy bear
teddy3.jpg	teddy bear
teddy4.jpg	teddy bear
teddy6.jpg	teddy bear
toaster.jpg	toaster
toaster2.png	toaster
toaster3.jpg	toaster

The training and testing images are located in the assets folders. These images belong to Wikimedia Commons.

Wikimedia Commons, the free media repository. Retrieved 10:48, October 17, 2018 from:
https://commons.wikimedia.org/wiki/Pizza
https://commons.wikimedia.org/wiki/Coffee_pot
https://commons.wikimedia.org/wiki/Toaster
https://commons.wikimedia.org/wiki/Category:Canoes
https://commons.wikimedia.org/wiki/Teddy_bear

Pre-trained model

There are multiple models which are pre-trained for classifying images. In this case, we will use a model based on an Inception topology, and trained with images from Image.Net. This model can be downloaded from https://storage.googleapis.com/download.tensorflow.org/models/inception5h.zip, but it's also available at / src / ImageClassification / assets /inputs / inception / tensorflow_inception_graph.pb.

Solution

The console application project ImageClassification.Score can be used to classify sample images based on the pre-trained Inception-5h TensorFlow model.

Again, note that this sample only uses/consumes a pre-trained TensorFlow model with ML.NET API. Therefore, it does not train any ML.NET model. Currently, TensorFlow is only supported in ML.NET for scoring/predicting with existing TensorFlow trained models.

You need to follow next steps in order to execute the classification test:

1. Set VS default startup project: Set ImageClassification.Score as starting project in Visual Studio.
2. Run the training model console app: Hit F5 in Visual Studio. At the end of the execution, the output will be similar to this screenshot:

Code Walkthrough

There is a single project in the solution named ImageClassification.Score, which is responsible for loading the model in TensorFlow format, and then classify images.

ML.NET: Model Scoring

Define the schema of data in a class type and refer that type while loading data using TextLoader. Here the class type is ImageNetData.

public class ImageNetData
{
    [LoadColumn(0)]
    public string ImagePath;

    [LoadColumn(1)]
    public string Label;

    public static IEnumerable<ImageNetData> ReadFromCsv(string file, string folder)
    {
        return File.ReadAllLines(file)
         .Select(x => x.Split('\t'))
         .Select(x => new ImageNetData()
         {
             ImagePath = Path.Combine(folder, x[0]),
             Label = x[1],
         });
    }
}

The first step is to load the data using TextLoader

var data = mlContext.Data.ReadFromTextFile<ImageNetData>(dataLocation, hasHeader: true);

The image file used to load images has two columns: the first one is defined as ImagePath and the second one is the Label corresponding to the image.

It is important to highlight that the label in the ImageNetData class is not really used when scoring with the TensorFlow model. It is used when testing the predictions so you can compare the actual label of each sample data with the predicted label provided by the TensorFlow model.

broccoli.jpg	broccoli
bucket.png	bucket
canoe.jpg	canoe
snail.jpg	snail
teddy1.jpg	teddy bear

As you can observe, the file does not have a header row.

The Inception model has several default parameters you need to pass in.

public struct ImageNetSettings
{
    public const int imageHeight = 224;
    public const int imageWidth = 224;
    public const float mean = 117;
    public const bool channelsLast = true;
}                

The second step is to define the estimator pipeline. Usually, when dealing with deep neural networks, you must adapt the images to the format expected by the network. This is the reason images are resized and then transformed (mainly, pixel values are normalized across all R,G,B channels).

var pipeline = 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(modelLocation)
        .ScoreTensorFlowModel(outputColumnNames: new[] { "softmax2" }, inputColumnNames: new[] { "input" },
            addBatchDimensionInput:true));

You also need to check the neural network, and check the names of the input / output nodes. In order to inspect the model, you can use tools like Netron, which is automatically installed with Visual Studio Tools for AI. These names are used later in the definition of the estimation pipe: in the case of the inception network, the input tensor is named 'input' and the output is named 'softmax2'

Finally, we extract the prediction engine after fitting the estimator pipeline. The prediction engine receives as parameter an object of type ImageNetData (containing 2 properties: ImagePath and Label), and then returns and object of type ImagePrediction.

ITransformer model = pipeline.Fit(data);
var predictionEngine = mlContext.Model.CreatePredictionEngine<ImageNetData, ImageNetPrediction>(model);

When obtaining the prediction, we get an array of floats in the property PredictedLabels. Each position in the array is assigned to a label, so for example, if the model has 5 different labels, the array will be length = 5. Each position in the array represents the label's probability in that position; the sum of all array values (probabilities) is equal to one. Then, you need to select the biggest value (probability), and check which is the assigned label to that position.

Citation

Training and prediction images

Wikimedia Commons, the free media repository. Retrieved 10:48, October 17, 2018 from https://commons.wikimedia.org/w/index.php?title=Main_Page&oldid=313158208.