add Prediction function for Transfer Learning

src/TensorFlowNET.Core/APIs/tf.graph.cs

@@ -24,7 +24,7 @@ namespace Tensorflow
             return ops.get_default_graph();
         }
 
-        public static Graph Graph() => new Graph();
-
+        public static Graph Graph() 
+            => new Graph();
     }
 }

src/TensorFlowNET.Core/Graphs/Graph.Import.cs

@@ -55,11 +55,11 @@ namespace Tensorflow
             return Status;
         }
 
-        public static Graph ImportFromPB(string file_path)
+        public static Graph ImportFromPB(string file_path, string name = null)
         {
             var graph = tf.Graph().as_default();
             var graph_def = GraphDef.Parser.ParseFrom(File.ReadAllBytes(file_path));
-            importer.import_graph_def(graph_def);
+            importer.import_graph_def(graph_def, name: name);
             return graph;
         }
     }

src/TensorFlowNET.Core/TensorFlowNET.Core.csproj

@@ -5,7 +5,7 @@
     <AssemblyName>TensorFlow.NET</AssemblyName>
     <RootNamespace>Tensorflow</RootNamespace>
     <TargetTensorFlow>1.14.0</TargetTensorFlow>
-    <Version>0.10.2</Version>
+    <Version>0.10.3</Version>
     <Authors>Haiping Chen, Meinrad Recheis</Authors>
     <Company>SciSharp STACK</Company>
     <GeneratePackageOnBuild>true</GeneratePackageOnBuild>
@@ -16,9 +16,8 @@
     <PackageIconUrl>https://avatars3.githubusercontent.com/u/44989469?s=200&amp;v=4</PackageIconUrl>
     <PackageTags>TensorFlow, NumSharp, SciSharp, MachineLearning, TensorFlow.NET, C#</PackageTags>
     <Description>Google's TensorFlow full binding in .NET Standard.
-Docs: https://tensorflownet.readthedocs.io
-Medium: https://medium.com/scisharp</Description>
-    <AssemblyVersion>0.10.2.0</AssemblyVersion>
+Docs: https://tensorflownet.readthedocs.io</Description>
+    <AssemblyVersion>0.10.3.0</AssemblyVersion>
     <PackageReleaseNotes>Changes since v0.9.0:
 
 1. Added full connected Convolution Neural Network example.
@@ -29,11 +28,12 @@ Medium: https://medium.com/scisharp</Description>
 6. Add StridedSliceGrad.
 7. Add BatchMatMulGrad.
 8. Upgrade NumSharp.
-9. Fix strided_slice_grad type convention error.</PackageReleaseNotes>
+9. Fix strided_slice_grad type convention error.
+10. Add AbsGrad.</PackageReleaseNotes>
     <LangVersion>7.2</LangVersion>
-    <FileVersion>0.10.2.0</FileVersion>
+    <FileVersion>0.10.3.0</FileVersion>
     <PackageLicenseFile>LICENSE</PackageLicenseFile>
-    <PackageRequireLicenseAcceptance>false</PackageRequireLicenseAcceptance>
+    <PackageRequireLicenseAcceptance>true</PackageRequireLicenseAcceptance>
     <SignAssembly>true</SignAssembly>
     <AssemblyOriginatorKeyFile>Open.snk</AssemblyOriginatorKeyFile>
   </PropertyGroup>

test/TensorFlowNET.Examples/ImageProcessing/RetrainImageClassifier.cs

@@ -80,6 +80,16 @@ namespace TensorFlowNET.Examples.ImageProcess
         {
             PrepareData();
 
+            #region For debug purpose
+            
+            // predict images
+            Predict(null);
+
+            // load saved pb and test new images.
+            Test(null); 
+            
+            #endregion
+
             var graph = IsImportingGraph ? ImportGraph() : BuildGraph();
 
             with(tf.Session(graph), sess =>
@@ -708,14 +718,38 @@ namespace TensorFlowNET.Examples.ImageProcess
             File.WriteAllText(output_labels, string.Join("\n", image_lists.Keys));
         }
 
-        public void Predict(Session sess)
+        public void Predict(Session sess_)
         {
-            throw new NotImplementedException();
+            if (!File.Exists(output_graph))
+                return;
+
+            var graph = Graph.ImportFromPB(output_graph, "");
+
+            Tensor input_layer = graph.OperationByName("input/BottleneckInputPlaceholder");
+            Tensor output_layer = graph.OperationByName("final_result");
+
+            with(tf.Session(graph), sess =>
+            {
+                // load images into NDArray in a matrix[image_num, features]
+                var nd = np.arange(2048f).reshape(1, 2048); // replace this line
+                var result = sess.run(output_layer, new FeedItem(input_layer, nd));
+            });
         }
 
-        public void Test(Session sess)
+        public void Test(Session sess_)
         {
-            throw new NotImplementedException();
+            if (!File.Exists(output_graph))
+                return;
+
+            var graph = Graph.ImportFromPB(output_graph);
+            var (jpeg_data_tensor, decoded_image_tensor) = add_jpeg_decoding();
+
+            with(tf.Session(graph), sess =>
+            {
+                (test_accuracy, predictions) = run_final_eval(sess, null, class_count, image_lists,
+                               jpeg_data_tensor, decoded_image_tensor, resized_image_tensor,
+                               bottleneck_tensor);
+            });
         }
     }
 }