Merge pull request #21 from microsoft/jay/merge-ic-sample

Merged ImageNet Classification models into one sample that they can all reference
microsoft · Nov 19, 2024 · e445cdd · e445cdd
2 parents 1b30ff7 + 3b6aa53
commit e445cdd
Show file tree

Hide file tree

Showing 6 changed files with 21 additions and 437 deletions.
diff --git a/...bileNet/ImageClassificationMobileNet.xaml → ... Models/ImageNet/ImageClassification.xaml b/...bileNet/ImageClassificationMobileNet.xaml → ... Models/ImageNet/ImageClassification.xaml
@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="utf-8" ?>
 <samples:BaseSamplePage
     xmlns:samples="using:AIDevGallery.Samples"
-    x:Class="AIDevGallery.Samples.OpenSourceModels.ImageClassificationMobileNet"
+    x:Class="AIDevGallery.Samples.OpenSourceModels.ImageClassification"
     xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
     xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
     xmlns:d="http://schemas.microsoft.com/expression/blend/2008"

diff --git a/...eNet/ImageClassificationMobileNet.xaml.cs → ...dels/ImageNet/ImageClassification.xaml.cs b/...eNet/ImageClassificationMobileNet.xaml.cs → ...dels/ImageNet/ImageClassification.xaml.cs
@@ -18,7 +18,7 @@
 namespace AIDevGallery.Samples.OpenSourceModels
 {
     [GallerySample(
-        Model1Types = [ModelType.MobileNet],
+        Model1Types = [ModelType.MobileNet, ModelType.ResNet, ModelType.SqueezeNet],
         Scenario = ScenarioType.ImageClassifyImage,
         NugetPackageReferences = [
             "System.Drawing.Common",
@@ -32,14 +32,14 @@ namespace AIDevGallery.Samples.OpenSourceModels
             SharedCodeEnum.BitmapFunctions,
             SharedCodeEnum.DeviceUtils
         ],
-        Name = "MobileNet Image Classification",
+        Name = "ImageNet Image Classification",
         Id = "09d73ba7-b877-45f9-9de6-41898ab4d339",
         Icon = "\uE8B9")]
-    internal sealed partial class ImageClassificationMobileNet : BaseSamplePage
+    internal sealed partial class ImageClassification : BaseSamplePage
     {
         private InferenceSession? _inferenceSession;
 
-        public ImageClassificationMobileNet()
+        public ImageClassification()
         {
             this.Unloaded += (s, e) => _inferenceSession?.Dispose();
             this.Loaded += (s, e) => Page_Loaded(); // <exclude-line>
@@ -102,11 +102,23 @@ private async void UploadImageButton_Click(object sender, RoutedEventArgs e)
 
         private async Task ClassifyImage(string filePath)
         {
-            if (!Path.Exists(filePath))
+            if (!Path.Exists(filePath) || _inferenceSession == null)
             {
                 return;
             }
 
+            // Grab model metadata
+            var inputName = _inferenceSession.InputNames[0];
+            var inputMetadata = _inferenceSession.InputMetadata[inputName];
+            var dimensions = inputMetadata.Dimensions;
+
+            // Set batch size to 1
+            int batchSize = 1;
+            dimensions[0] = batchSize;
+
+            int inputWidth = dimensions[2];
+            int inputHeight = dimensions[3];
+
             BitmapImage bitmapImage = new(new Uri(filePath));
             UploadedImage.Source = bitmapImage;
             NarratorHelper.AnnounceImageChanged(UploadedImage, "Image changed: new upload."); // <exclude-line>
@@ -116,22 +128,19 @@ private async Task ClassifyImage(string filePath)
                 Bitmap image = new(filePath);
 
                 // Resize image
-                int width = 224;
-                int height = 224;
-                var resizedImage = BitmapFunctions.ResizeBitmap(image, width, height);
+                var resizedImage = BitmapFunctions.ResizeBitmap(image, inputWidth, inputHeight);
                 image.Dispose();
                 image = resizedImage;
 
                 // Preprocess image
-                Tensor<float> input = new DenseTensor<float>([1, 3, 224, 224]);
+                Tensor<float> input = new DenseTensor<float>(dimensions);
                 input = BitmapFunctions.PreprocessBitmapWithStdDev(image, input);
                 image.Dispose();
 
                 // Setup inputs
-                var inputMetadataName = _inferenceSession!.InputNames[0];
                 var inputs = new List<NamedOnnxValue>
                 {
-                    NamedOnnxValue.CreateFromTensor(inputMetadataName, input)
+                    NamedOnnxValue.CreateFromTensor(inputName, input)
                 };
 
                 // Run inference

diff --git a/AIDevGallery/Samples/Open Source Models/Image Models/ResNet/ImageClassificationResNet.xaml b/AIDevGallery/Samples/Open Source Models/Image Models/ResNet/ImageClassificationResNet.xaml
diff --git a/...vGallery/Samples/Open Source Models/Image Models/ResNet/ImageClassificationResNet.xaml.cs b/...vGallery/Samples/Open Source Models/Image Models/ResNet/ImageClassificationResNet.xaml.cs
diff --git a/...ery/Samples/Open Source Models/Image Models/SqueezeNet/ImageClassificationSqueezeNet.xaml b/...ery/Samples/Open Source Models/Image Models/SqueezeNet/ImageClassificationSqueezeNet.xaml