image classification #23

SciSharp · Dec 19, 2020 · 0316fde · 0316fde
1 parent 43c06d5
commit 0316fde
Show file tree

Hide file tree

Showing 8 changed files with 135 additions and 43 deletions.
diff --git a/README.md b/README.md
@@ -1,5 +1,5 @@
 # SciSharp STACK Examples
-This repo contains many practical examples written in SciSharp's machine learning libraries. If you still don't know how to use .NET for deep learning, getting started from here is your best choice.
+This repo contains many practical examples written in SciSharp's machine learning libraries. If you still don't know how to use .NET for deep learning, getting started from these examples is your best choice.
 
 [![Join the chat at https://gitter.im/publiclab/publiclab](https://badges.gitter.im/Join%20Chat.svg)](https://gitter.im/sci-sharp/community)
 
@@ -30,6 +30,8 @@ dotnet TensorFlowNET.Examples.dll -ex "MNIST CNN (Eager)"
 
 Example runner will download all the required files like training data and model pb files.
 
+#### Basic Model
+
 * Hello World [C#](src/TensorFlowNET.Examples/HelloWorld.cs)
 * Basic Operations [C#](src/TensorFlowNET.Examples/BasicOperations.cs)
 * Linear Regression in Graph mode [C#](src/TensorFlowNET.Examples/BasicModels/LinearRegression.cs) 
@@ -38,22 +40,29 @@ Example runner will download all the required files like training data and model
 * Logistic Regression in Eager mode [C#](src/TensorFlowNET.Examples/BasicModels/LogisticRegressionEager.cs) 
 * Nearest Neighbor [C#](src/TensorFlowNET.Examples/BasicModels/NearestNeighbor.cs)
 * Naive Bayes Classification [C#](src/TensorFlowNET.Examples/BasicModels/NaiveBayesClassifier.cs)
-* Full Connected Neural Network in Eager mode [C#](src/TensorFlowNET.Examples/\NeuralNetworks/FullyConnectedEager.cs) 
 * K-means Clustering [C#](src/TensorFlowNET.Examples/BasicModels/KMeansClustering.cs)
+
+#### Neural Network
+
+* Full Connected Neural Network in Eager mode [C#](src/TensorFlowNET.Examples/\NeuralNetworks/FullyConnectedEager.cs) 
 * NN XOR [C#](src/TensorFlowNET.Examples/NeuralNetworks/NeuralNetXor.cs)
 * Object Detection in MobileNet [C#](src/TensorFlowNET.Examples/ObjectDetection/DetectInMobilenet.cs) 
-* Binary Text Classification [C#](src/TensorFlowNET.Examples/TextProcessing/BinaryTextClassification.cs)
-* CNN Text Classification [C#](src/TensorFlowNET.Examples/TextProcessing/cnn_models/VdCnn.cs)
 * MNIST FNN in Keras Functional API [C#](src/TensorFlowNET.Examples/ImageProcessing/MnistFnnKerasFunctional.cs) 
 * MNIST CNN in Graph mode [C#](src/TensorFlowNET.Examples/ImageProcessing/DigitRecognitionCNN.cs) 
 * MNIST RNN [C#](src/TensorFlowNET.Examples/ImageProcessing/DigitRecognitionRNN.cs)
 * MNIST LSTM [C#](src/TensorFlowNET.Examples/ImageProcessing/DigitRecognitionLSTM.cs)
-* Named Entity Recognition [C#](src/TensorFlowNET.Examples/TextProcessing/NER)
+* Image Classification in Keras Sequential API [C#](src/TensorFlowNET.Examples/ImageProcessing/ImageClassificationKeras.cs)
+* Toy ResNet in Keras Functional API [C#](src/TensorFlowNET.Examples/ImageProcessing/ToyResNet.cs)
 * Transfer Learning for Image Classification in InceptionV3 [C#](src/TensorFlowNET.Examples/ImageProcessing/TransferLearningWithInceptionV3.cs)
 * CNN In Your Own Dataset [C#](src/TensorFlowNET.Examples/ImageProcessing/CnnInYourOwnData.cs)
 
+#### Natural Language Processing
+* Binary Text Classification [C#](src/TensorFlowNET.Examples/TextProcessing/BinaryTextClassification.cs)
+* CNN Text Classification [C#](src/TensorFlowNET.Examples/TextProcessing/cnn_models/VdCnn.cs)
+* Named Entity Recognition [C#](src/TensorFlowNET.Examples/TextProcessing/NER)
+
 
-### Welcome to PR your example to us.
+#### Welcome to PR your example to us.
 Your contribution will make .NET community better than ever.
 <br>
 <a href="http://scisharpstack.org"><img src="https://github.com/SciSharp/SciSharp/blob/master/art/scisharp-stack.png" width="391" height="100" /></a>
diff --git a/src/TensorFlowNET.Examples/ImageProcessing/ImageClassificationKeras.cs b/src/TensorFlowNET.Examples/ImageProcessing/ImageClassificationKeras.cs
@@ -3,6 +3,10 @@
 using Tensorflow.Keras;
 using static Tensorflow.Binding;
 using static Tensorflow.KerasApi;
+using System.Linq;
+using Tensorflow.Keras.Utils;
+using System.IO;
+using Tensorflow.Keras.Engine;
 
 namespace TensorFlowNET.Examples
 {
@@ -12,67 +16,100 @@ namespace TensorFlowNET.Examples
     /// </summary>
     public class ImageClassificationKeras : SciSharpExample, IExample
     {
+        int batch_size = 32;
+        int epochs = 10;
+        TensorShape img_dim = (180, 180);
+        IDatasetV2 train_ds, val_ds;
+        Model model;
+
         public ExampleConfig InitConfig()
             => Config = new ExampleConfig
             {
                 Name = "Image Classification (Keras)",
-                Enabled = false,
+                Enabled = true,
                 Priority = 18
             };
 
         public bool Run()
         {
+            tf.enable_eager_execution();
+
             PrepareData();
+            BuildModel();
+            Train();
+
             return true;
         }
 
+        public override void BuildModel()
+        {
+            int num_classes = 5;
+            // var normalization_layer = tf.keras.layers.Rescaling(1.0f / 255);
+            var layers = keras.layers;
+            model = keras.Sequential(new List<ILayer>
+            {
+                layers.Rescaling(1.0f / 255, input_shape: (img_dim.dims[0], img_dim.dims[1], 3)),
+                layers.Conv2D(16, 3, padding: "same", activation: keras.activations.Relu),
+                layers.MaxPooling2D(),
+                /*layers.Conv2D(32, 3, padding: "same", activation: "relu"),
+                layers.MaxPooling2D(),
+                layers.Conv2D(64, 3, padding: "same", activation: "relu"),
+                layers.MaxPooling2D(),*/
+                layers.Flatten(),
+                layers.Dense(128, activation: keras.activations.Relu),
+                layers.Dense(num_classes)
+            });
+
+            model.compile(optimizer: keras.optimizers.Adam(),
+                loss: keras.losses.SparseCategoricalCrossentropy(from_logits: true),
+                metrics: new[] { "accuracy" });
+
+            model.summary();
+        }
+
+        public override void Train()
+        {
+            model.fit(train_ds, validation_data: val_ds, epochs: epochs);
+        }
+
         public override void PrepareData()
         {
-            int batch_size = 32;
-            TensorShape img_dim = (180, 180);
+            string fileName = "flower_photos.tgz";
+            string url = $"https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz";
+            string data_dir = Path.GetTempPath();
+            Web.Download(url, data_dir, fileName);
+            Compress.ExtractTGZ(Path.Join(data_dir, fileName), data_dir);
+            data_dir = Path.Combine(data_dir, "flower_photos");
 
-            var data_dir = @"C:/Users/haipi/.keras/datasets/flower_photos";
-            var train_ds = keras.preprocessing.image_dataset_from_directory(data_dir,
+            // convert to tensor
+            train_ds = keras.preprocessing.image_dataset_from_directory(data_dir,
                 validation_split: 0.2f,
                 subset: "training",
                 seed: 123,
                 image_size: img_dim,
                 batch_size: batch_size);
 
-            var val_ds = keras.preprocessing.image_dataset_from_directory(data_dir,
+            val_ds = keras.preprocessing.image_dataset_from_directory(data_dir,
                 validation_split: 0.2f,
                 subset: "validation",
                 seed: 123,
                 image_size: img_dim,
                 batch_size: batch_size);
 
-            train_ds = train_ds.cache().shuffle(100).prefetch(buffer_size: -1);
+            train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size: -1);
             val_ds = val_ds.cache().prefetch(buffer_size: -1);
 
             foreach (var (img, label) in train_ds)
             {
-                print("batch images: " + img.TensorShape);
-                print("labels: " + label);
+                print($"images: {img.TensorShape}");
+                var nd = label.numpy();
+                print($"labels: {nd}");
+                var data = nd.Data<int>();
+                if (data.Max() > 4 || data.Min() < 0)
+                {
+                    // exception
+                }
             }
-
-            int num_classes = 5;
-            // var normalization_layer = tf.keras.layers.Rescaling(1.0f / 255);
-            var layers = keras.layers;
-            var model = keras.Sequential(new List<ILayer>
-            {
-                layers.Rescaling(1.0f / 255, input_shape: (img_dim.dims[0], img_dim.dims[1], 3)),
-                layers.Conv2D(16, 3, padding: "same", activation: keras.activations.Relu),
-                layers.MaxPooling2D(),
-                /*layers.Conv2D(32, 3, padding: "same", activation: "relu"),
-                layers.MaxPooling2D(),
-                layers.Conv2D(64, 3, padding: "same", activation: "relu"),
-                layers.MaxPooling2D(),*/
-                layers.Flatten(),
-                layers.Dense(128, activation: keras.activations.Relu),
-                layers.Dense(num_classes)
-            });
-
-            model.compile("adam", keras.losses.SparseCategoricalCrossentropy(from_logits: true));
         }
     }
 }
diff --git a/src/TensorFlowNET.Examples/TensorFlowNET.Examples.csproj b/src/TensorFlowNET.Examples/TensorFlowNET.Examples.csproj
@@ -9,7 +9,7 @@
 
   <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|AnyCPU'">
     <DefineConstants>DEBUG;TRACE</DefineConstants>
-    <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
+    <AllowUnsafeBlocks>false</AllowUnsafeBlocks>
     <PlatformTarget>AnyCPU</PlatformTarget>
   </PropertyGroup>
 
@@ -38,13 +38,12 @@
   </ItemGroup>
 
   <ItemGroup>
-    <PackageReference Include="Colorful.Console" Version="1.2.11" />
+    <PackageReference Include="Colorful.Console" Version="1.2.15" />
     <PackageReference Include="Newtonsoft.Json" Version="12.0.3" />
     <PackageReference Include="OpenCvSharp4.runtime.win" Version="4.4.0.20200915" />
     <PackageReference Include="SciSharp.TensorFlow.Redist" Version="2.3.1" />
     <PackageReference Include="SharpCV" Version="0.6.0" />
-    <PackageReference Include="System.Drawing.Common" Version="4.7.0" />
-    <PackageReference Include="TensorFlow.Keras" Version="0.2.1" />
+    <PackageReference Include="System.Drawing.Common" Version="5.0.0" />
   </ItemGroup>
 
   <ItemGroup>

diff --git a/src/TensorFlowNET.Examples/TextProcessing/Word2Vec.cs b/src/TensorFlowNET.Examples/TextProcessing/Word2Vec.cs
@@ -50,6 +50,8 @@ public ExampleConfig InitConfig()
 
         public bool Run()
         {
+            tf.compat.v1.disable_eager_execution();
+
             PrepareData();
 
             var graph = tf.Graph().as_default();

diff --git a/src/tensorflow2.x-python-tutorial/.vscode/launch.json b/src/tensorflow2.x-python-tutorial/.vscode/launch.json
@@ -0,0 +1,17 @@
+{
+  // Use IntelliSense to learn about possible attributes.
+  // Hover to view descriptions of existing attributes.
+  // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
+  "version": "0.2.0",
+  "configurations": [
+    {
+      "name": "Python: Current File",
+      "type": "python",
+      "request": "launch",
+      "program": "${workspaceFolder}/image_classification.py",
+      "console": "integratedTerminal",
+      "cwd": "${workspaceFolder}",
+      "justMyCode": false
+    }
+  ]
+}
diff --git a/src/tensorflow2.x-python-tutorial/MapDataset.py b/src/tensorflow2.x-python-tutorial/MapDataset.py
@@ -2,14 +2,17 @@
 
 # Import TensorFlow v2.
 import tensorflow as tf
+from tensorflow.python.data.experimental.ops import cardinality
 
 tf.autograph.set_verbosity(10, alsologtostdout=True)
+tf.config.run_functions_eagerly(True)
 
 def add(x):
     print("debug test test")
     return x + 1
 
-dataset = tf.data.Dataset.range(1, 3)  # ==> [ 1, 2, 3, 4, 5 ]
+dataset = tf.data.Dataset.range(10)  # ==> [ 1, 2, 3, 4, 5 ]
 dataset = dataset.map(add)
+card = cardinality.cardinality(dataset)
 for item in dataset:
-    print(item)
+    print(item)
diff --git a/src/tensorflow2.x-python-tutorial/image_classification.py b/src/tensorflow2.x-python-tutorial/image_classification.py
@@ -12,6 +12,13 @@
 from tensorflow.keras.models import Sequential
 from tensorflow.python.ops import bitwise_ops
 
+"""
+y_true = np.array([[0., 1.], [0., 0.]],dtype='float64')
+y_pred = np.array([[1., 1.], [1., 0.]],dtype='float64')
+mse = tf.keras.losses.MeanSquaredError()
+print(mse(y_true, y_pred).numpy())
+"""
+
 import pathlib
 dataset_url = "https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz"
 data_dir = tf.keras.utils.get_file('flower_photos', origin=dataset_url, untar=True)
@@ -24,14 +31,21 @@
 img_height = 180
 img_width = 180
 
+tf.config.run_functions_eagerly(True)
+
 train_ds = tf.keras.preprocessing.image_dataset_from_directory(
   data_dir,
   validation_split=0.2,
   subset="training",
   seed=123,
+  shuffle=False,
   image_size=(img_height, img_width),
   batch_size=batch_size)
 
+for img, label in train_ds:     
+  print("batch images: ", img.shape)
+  print("labels: ", label)
+
 val_ds = tf.keras.preprocessing.image_dataset_from_directory(
   data_dir,
   validation_split=0.2,
@@ -47,7 +61,7 @@
 
 train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)
 val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)
-
+            
 normalization_layer = layers.experimental.preprocessing.Rescaling(1./255)
 
 normalized_ds = train_ds.map(lambda x, y: (normalization_layer(x), y))
@@ -82,8 +96,7 @@
 history = model.fit(
   train_ds,
   validation_data=val_ds,
-  epochs=epochs
-)
+  epochs=epochs)
 
 acc = history.history['accuracy']
 val_acc = history.history['val_accuracy']

diff --git a/src/tensorflow2.x-python-tutorial/text.py b/src/tensorflow2.x-python-tutorial/text.py
@@ -0,0 +1,12 @@
+import numpy as np
+import tensorflow as tf
+import tensorflow_text as text
+
+docs = tf.constant([u'Everything not saved will be lost.'])
+tokenizer = text.WhitespaceTokenizer()
+tokens = tokenizer.tokenize(docs)
+f1 = text.wordshape(tokens, text.WordShape.HAS_TITLE_CASE)
+bigrams = text.ngrams(tokens, 2, reduction_type=text.Reduction.STRING_JOIN)
+
+print(docs)
+a = input()