Image classification using DNNs and Transfer Learning.

docs/samples/Microsoft.ML.Samples/Dynamic/TensorFlow/ImageClassification.cs

@@ -71,7 +71,7 @@ public static void Example()
             //----------
         }
 
-        private const int imageHeight = 224; 
+        private const int imageHeight = 224;
         private const int imageWidth = 224;
         private const int numChannels = 3;
         private const int inputSize = imageHeight * imageWidth * numChannels;
@@ -95,7 +95,7 @@ public static TensorData[] GetTensorData()
             // This can be any numerical data. Assume image pixel values.
             var image1 = Enumerable.Range(0, inputSize).Select(
                 x => (float)x / inputSize).ToArray();
-            
+
             var image2 = Enumerable.Range(0, inputSize).Select(
                 x => (float)(x + 10000) / inputSize).ToArray();
             return new TensorData[] { new TensorData() { input = image1 },
@@ -137,4 +137,4 @@ private static void Unzip(string path, string targetDir)
             inStream.Close();
         }
     }
-}
+}

docs/samples/Microsoft.ML.Samples/Dynamic/TensorFlow/TextClassification.cs

@@ -55,9 +55,8 @@ public static void Example()
             //      - Use it for quering the schema for input and output in the
             //            model
             //      - Use it for prediction in the pipeline.
-            var tensorFlowModel = mlContext.Model.LoadTensorFlowModel(
-                modelLocation);
-            var schema = tensorFlowModel.GetModelSchema();
+            var tensorFlowModel = mlContext.Model.LoadTensorFlowModel(modelLocation);
+            /*var schema = tensorFlowModel.GetModelSchema();
             var featuresType = (VectorDataViewType)schema["Features"].Type;
             Console.WriteLine("Name: {0}, Type: {1}, Shape: (-1, {2})", "Features",
                 featuresType.ItemType.RawType, featuresType.Dimensions[0]);
@@ -66,7 +65,7 @@ public static void Example()
                 .Type;
             Console.WriteLine("Name: {0}, Type: {1}, Shape: (-1, {2})",
                 "Prediction/Softmax", predictionType.ItemType.RawType,
-                predictionType.Dimensions[0]);
+                predictionType.Dimensions[0]);*/
 
             // The model expects the input feature vector to be a fixed length
             // vector.

docs/samples/Microsoft.ML.Samples/Dynamic/TensorFlow/TransferLearning.cs

@@ -0,0 +1,140 @@
+using System;
+using System.Diagnostics;
+using System.IO;
+using System.Linq;
+using System.Net;
+using ICSharpCode.SharpZipLib.GZip;
+using ICSharpCode.SharpZipLib.Tar;
+using Microsoft.ML;
+using Microsoft.ML.Data;
+
+namespace Samples.Dynamic
+{
+    public static class TransferLearning
+    {
+        /// <summary>
+        /// Example use of the TensorFlow image model in a ML.NET pipeline.
+        /// </summary>
+        public static void Example()
+        {
+            var sw = new Stopwatch();
+
+            var mlContext = new MLContext();
+            var data = GetTensorData();
+            var idv = mlContext.Data.LoadFromEnumerable(data);
+
+            // Create a ML pipeline.
+            var pipeline = 
+                mlContext.Transforms.Conversion.MapValueToKey(
+                    nameof(TensorData.Label))
+                .Append(mlContext.Model.ImageClassification(
+                    nameof(TensorData.input),
+                    nameof(TensorData.Label), batchSize: 2,
+                        addBatchDimensionInput: true));
+
+            // Run the pipeline and get the transformed values.
+            var estimator = pipeline.Fit(idv);
+            var transformedValues = estimator.Transform(idv);
+
+            // Retrieve model scores.
+            var outScores = mlContext.Data.CreateEnumerable<OutputScores>(
+                transformedValues, reuseRowObject: false);
+
+            // Display scores. (for the sake of brevity we display scores of the
+            // first 3 classes)
+            foreach (var prediction in outScores)
+            {
+                int numClasses = 0;
+                foreach (var classScore in prediction.Scores.Take(2))
+                {
+                    Console.WriteLine(
+                        $"Class #{numClasses++} score = {classScore}");
+                }
+                Console.WriteLine(
+                    $"Predicted Label: {prediction.PredictedLabel}");
+
+
+                Console.WriteLine(new string('-', 10));
+            }
+
+            Console.WriteLine(sw.Elapsed);
+            // Class #0 score = 0.03416626
+            // Class #1 score = 0.9658337
+            // Predicted Label: 1
+            // ----------
+            // Class #0 score = 0.02959618
+            // Class #1 score = 0.9704038
+            // Predicted Label: 1
+            // ----------
+        }
+
+        private const int imageHeight = 224; 
+        private const int imageWidth = 224;
+        private const int numChannels = 3;
+        private const int inputSize = imageHeight * imageWidth * numChannels;
+
+        /// <summary>
+        /// A class to hold sample tensor data. 
+        /// Member name should match the inputs that the model expects (in this
+        /// case, input).
+        /// </summary>
+        public class TensorData
+        {
+            [VectorType(imageHeight, imageWidth, numChannels)]
+            public float[] input { get; set; }
+
+            public Int64 Label { get; set; }
+        }
+
+        /// <summary>
+        /// Method to generate sample test data. Returns 2 sample rows.
+        /// </summary>
+        public static TensorData[] GetTensorData()
+        {
+            // This can be any numerical data. Assume image pixel values.
+            var image1 = Enumerable.Range(0, inputSize).Select(
+                x => (float)x / inputSize).ToArray();
+
+            var image2 = Enumerable.Range(0, inputSize).Select(
+                x => (float)(x + 10000) / inputSize).ToArray();
+            return new TensorData[] { new TensorData() { input = image1, Label = 0 },
+                new TensorData() { input = image2, Label = 1 } };
+        }
+
+        /// <summary>
+        /// Class to contain the output values from the transformation.
+        /// </summary>
+        class OutputScores
+        {
+            public float[] Scores { get; set; }
+            public Int64 PredictedLabel { get; set; }
+        }
+
+        private static string Download(string baseGitPath, string dataFile)
+        {
+            using (WebClient client = new WebClient())
+            {
+                client.DownloadFile(new Uri($"{baseGitPath}"), dataFile);
+            }
+
+            return dataFile;
+        }
+
+        /// <summary>
+        /// Taken from 
+        /// https://github.com/icsharpcode/SharpZipLib/wiki/GZip-and-Tar-Samples.
+        /// </summary>
+        private static void Unzip(string path, string targetDir)
+        {
+            Stream inStream = File.OpenRead(path);
+            Stream gzipStream = new GZipInputStream(inStream);
+
+            TarArchive tarArchive = TarArchive.CreateInputTarArchive(gzipStream);
+            tarArchive.ExtractContents(targetDir);
+            tarArchive.Close();
+
+            gzipStream.Close();
+            inStream.Close();
+        }
+    }
+}

docs/samples/Microsoft.ML.Samples/Microsoft.ML.Samples.csproj

@@ -10,6 +10,7 @@
   </PropertyGroup>
 
   <ItemGroup>
+	<ProjectReference Include="..\..\..\src\Microsoft.ML.Dnn\Microsoft.ML.Dnn.csproj" />
     <ProjectReference Include="..\..\..\src\Microsoft.ML.LightGbm\Microsoft.ML.LightGbm.csproj" />
     <ProjectReference Include="..\..\..\src\Microsoft.ML.Mkl.Components\Microsoft.ML.Mkl.Components.csproj" />
     <ProjectReference Include="..\..\..\src\Microsoft.ML.KMeansClustering\Microsoft.ML.KMeansClustering.csproj" />
@@ -963,5 +964,10 @@
       <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
     </Content>
   </ItemGroup>
-
+    <ItemGroup>
+    <Content Include="$(ObjDir)DnnImageModels\Resnet101V2Tensorflow\resnet_v2_101_299.meta">
+      <Link>DnnImageModels\Resnet101V2Tensorflow\resnet_v2_101_299.meta</Link>
+      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
+    </Content>
+  </ItemGroup>
 </Project>

docs/samples/Microsoft.ML.Samples/Program.cs

@@ -10,7 +10,7 @@ public static class Program
 
         internal static void RunAll()
         {
-            int samples = 0;
+            /*int samples = 0;
             foreach (var type in Assembly.GetExecutingAssembly().GetTypes())
             {
                 var sample = type.GetMethod("Example", BindingFlags.Public | BindingFlags.Static | BindingFlags.FlattenHierarchy);
@@ -23,7 +23,8 @@ internal static void RunAll()
                 }
             }
 
-            Console.WriteLine("Number of samples that ran without any exception: " + samples);
+            Console.WriteLine("Number of samples that ran without any exception: " + samples);*/
+            TransferLearning.Example();
         }
     }
 }

pkg/Microsoft.ML.Dnn/Microsoft.ML.Dnn.nupkgproj

@@ -0,0 +1,12 @@
+<Project Sdk="Microsoft.NET.Sdk" DefaultTargets="Pack">
+
+  <PropertyGroup>
+    <TargetFramework>netstandard2.0</TargetFramework>
+    <PackageDescription>Microsoft.ML.Dnn contains APIs to do high level DNN training such as image classification.</PackageDescription>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="../Microsoft.ML/Microsoft.ML.nupkgproj" />
+  </ItemGroup>
+
+</Project>

pkg/Microsoft.ML.Dnn/Microsoft.ML.Dnn.symbols.nupkgproj

@@ -0,0 +1,5 @@
+<Project DefaultTargets="Pack">
+
+  <Import Project="Microsoft.ML.Dnn.nupkgproj" />
+
+</Project>

pkg/common/Resnet101V2.props

@@ -0,0 +1,9 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <Content Include="$(MSBuildThisFileDirectory)..\..\tools\DnnImageModels\Resnet101V2Tensorflow\*.*">
+      <Link>DnnImageModels\%(RecursiveDir)%(Filename)%(Extension)</Link>
+      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
+    </Content>
+  </ItemGroup>
+</Project>

src/Microsoft.ML.Core/Properties/AssemblyInfo.cs

@@ -37,6 +37,7 @@
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.StandardTrainers" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Sweeper" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.TensorFlow" + PublicKey.Value)]
+[assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Dnn" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.TimeSeries" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Transforms" + PublicKey.Value)]
 

src/Microsoft.ML.Data/MLContext.cs

@@ -110,7 +110,7 @@ public sealed class MLContext : IHostEnvironment
         public MLContext(int? seed = null)
         {
             _env = new LocalEnvironment(seed);
-            _env.AddListener(ProcessMessage);
+            _env.AddListener<ChannelMessage>(ProcessMessage);
 
             BinaryClassification = new BinaryClassificationCatalog(_env);
             MulticlassClassification = new MulticlassClassificationCatalog(_env);

src/Microsoft.ML.Data/Properties/AssemblyInfo.cs

@@ -36,6 +36,7 @@
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.StandardTrainers" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Sweeper" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.TensorFlow" + PublicKey.Value)]
+[assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Dnn" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.TimeSeries" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Transforms" + PublicKey.Value)]
 [assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.DnnImageFeaturizer.AlexNet" + PublicKey.Value)]

src/Microsoft.ML.Dnn/AssemblyInfo.cs

@@ -0,0 +1,9 @@
+// Licensed to the .NET Foundation under one or more agreements.
+// The .NET Foundation licenses this file to you under the MIT license.
+// See the LICENSE file in the project root for more information.
+
+using System.Runtime.CompilerServices;
+using Microsoft.ML;
+
+[assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Tests" + PublicKey.TestValue)]
+[assembly: InternalsVisibleTo(assemblyName: "Microsoft.ML.Tensorflow" + PublicKey.Value)]