Add result text and fix few issues in code

handwriting/handwriting.proto

@@ -44,5 +44,6 @@ message None {}
 service Handwriting {
     rpc recognize(Matrix) returns (Result) {}
     rpc setNeuralNetworkData(NeuralNetworkRaw) returns (None) {}
+    rpc getNeuralNetworkData(None) returns (NeuralNetworkRaw) {}
     rpc reTrain(None) returns (None) {}
 }

handwriting/handwriting/handwriting.go

@@ -4,7 +4,9 @@ import (
 	"bytes"
 	context "context"
 	fmt "fmt"
+	"log"
 	"net"
+	"os"
 
 	neuralnetwork "../../neuralnetwork/neuralnetwork"
 	training "../../neuralnetwork/training"
@@ -42,11 +44,41 @@ func (hws *HandwritingService) SetNeuralNetworkData(ctx context.Context, nnRaw *
 	return &None{}, nil
 }
 
+func (hws *HandwritingService) GetNeuralNetworkData(context.Context, *None) (*NeuralNetworkRaw, error) {
+	nnRaw := &NeuralNetworkRaw{}
+	fmt.Println("SetNeuralNetworkData")
+	r := bytes.NewReader(nnRaw.Data)
+	hws.nn.LoadState(r)
+	return nnRaw, nil
+}
+
 func (hws *HandwritingService) ReTrain(context.Context, *None) (*None, error) {
 	fmt.Println("ReTrain")
+
 	trainer := training.NewMNISTReader("./mnist.data", "./mnist.labels")
+	failCount := 0
+	total := 0
+	trainer.Reset()
+	for trainer.NextValidator() {
+		total++
+		dataSet, expect := trainer.GetValidator()
+		index, _ := hws.nn.Predict(dataSet)
+		if expect.At(index, 0) != 1.0 {
+			failCount++
+			// fmt.Printf("Fail: %v, %v\n\n", trainer.ValidationIndex(), expect.At(index, 0))
+		}
+	}
+	fmt.Printf("Fail count: %v/%v\n\n", failCount, total)
+
 	hws.nn.Train(trainer, 100)
+
 	fmt.Println("ReTrain finished")
+	outFile, err := os.OpenFile("./mnistnet.nnd", os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0666)
+	if err != nil {
+		log.Fatal(err)
+	}
+	defer outFile.Close()
+	hws.nn.SaveState(outFile)
 	return &None{}, nil
 }
 
@@ -63,3 +95,16 @@ func (hws *HandwritingService) Run() {
 		fmt.Printf("Failed to serve: %v\n", err)
 	}
 }
+
+func drawImage(dense *mat.Dense) {
+	for i := 0; i < 28; i++ {
+		for j := 0; j < 28; j++ {
+			val := 0
+			if dense.At(i*28+j, 0) > 0 {
+				val = 1
+			}
+			fmt.Printf("%v ", val)
+		}
+		fmt.Println()
+	}
+}

handwriting/handwriting/handwriting.pb.go

@@ -182,22 +182,22 @@ func init() {
 func init() { proto.RegisterFile("handwriting.proto", fileDescriptor_d3287f4c1e120e43) }
 
 var fileDescriptor_d3287f4c1e120e43 = []byte{
-	// 228 bytes of a gzipped FileDescriptorProto
-	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0x6c, 0x90, 0xc1, 0x4a, 0x03, 0x41,
-	0x0c, 0x86, 0x3b, 0x58, 0x46, 0x4c, 0x15, 0x6d, 0xf4, 0x50, 0x8a, 0x42, 0x99, 0x83, 0xcc, 0xa9,
-	0x60, 0x3d, 0xf8, 0x00, 0x7a, 0x10, 0xc1, 0x3d, 0x0c, 0xbe, 0x40, 0x6c, 0x43, 0x3b, 0x58, 0x66,
-	0x24, 0x4d, 0x59, 0xf1, 0xfd, 0x7c, 0x2f, 0x61, 0x57, 0x70, 0x67, 0xf5, 0xf6, 0xf3, 0xff, 0xc9,
-	0x9f, 0x8f, 0xc0, 0x78, 0x43, 0x69, 0x55, 0x4b, 0xd4, 0x98, 0xd6, 0xf3, 0x77, 0xc9, 0x9a, 0x71,
-	0xd4, 0xb1, 0xdc, 0x25, 0xd8, 0x67, 0x52, 0x89, 0x1f, 0x88, 0x30, 0x5c, 0x91, 0xd2, 0xc4, 0xcc,
-	0x0e, 0xbc, 0x09, 0x8d, 0x76, 0x0b, 0xb0, 0x81, 0x77, 0xfb, 0xad, 0xa2, 0x87, 0x53, 0x69, 0xd4,
-	0xfd, 0x86, 0x84, 0x96, 0xca, 0x32, 0x31, 0x33, 0xe3, 0x4f, 0x42, 0xdf, 0x76, 0xd7, 0x70, 0x56,
-	0xf1, 0x5e, 0x68, 0x5b, 0xb1, 0xd6, 0x59, 0xde, 0x02, 0xd5, 0x9d, 0x6e, 0xe3, 0x8f, 0x7f, 0xba,
-	0x2d, 0x0c, 0xab, 0x9c, 0x78, 0xf1, 0x65, 0x60, 0xf4, 0xf8, 0x4b, 0x84, 0x77, 0x70, 0x24, 0xbc,
-	0xcc, 0xeb, 0x14, 0x3f, 0x19, 0xcf, 0xe7, 0x5d, 0xfe, 0x96, 0x74, 0x5a, 0x9a, 0x2d, 0xa0, 0x1b,
-	0xe0, 0x13, 0x5c, 0xec, 0x58, 0x8b, 0xdb, 0x0f, 0xa4, 0x84, 0x57, 0xc5, 0x78, 0x9f, 0x6d, 0x3a,
-	0x2e, 0xe3, 0x9c, 0xd8, 0x0d, 0xf0, 0x06, 0x0e, 0x85, 0x5f, 0x84, 0x62, 0xc2, 0xbf, 0xf9, 0xbf,
-	0x2b, 0xaf, 0xb6, 0xf9, 0xee, 0xed, 0x77, 0x00, 0x00, 0x00, 0xff, 0xff, 0x46, 0x56, 0x5b, 0x3e,
-	0x72, 0x01, 0x00, 0x00,
+	// 238 bytes of a gzipped FileDescriptorProto
+	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0x7c, 0x91, 0xc1, 0x4a, 0xc3, 0x40,
+	0x10, 0x86, 0xbb, 0x5a, 0x22, 0x4e, 0x15, 0xed, 0xe8, 0xa1, 0x04, 0x85, 0xb2, 0x07, 0xc9, 0xa9,
+	0x60, 0x3d, 0xf8, 0x00, 0x7a, 0x10, 0xc1, 0x1c, 0x16, 0x5f, 0x60, 0x6c, 0x87, 0x74, 0xb1, 0xec,
+	0xca, 0x74, 0x4a, 0xc4, 0xb7, 0xf0, 0x8d, 0x85, 0x44, 0x30, 0x5b, 0x43, 0x6f, 0xc3, 0xbf, 0xb3,
+	0xdf, 0xff, 0xc1, 0xc0, 0x78, 0x45, 0x61, 0x59, 0x8b, 0x57, 0x1f, 0xaa, 0xd9, 0x87, 0x44, 0x8d,
+	0x38, 0xea, 0x44, 0xf6, 0x0a, 0xb2, 0x17, 0x52, 0xf1, 0x9f, 0x88, 0x30, 0x5c, 0x92, 0xd2, 0xc4,
+	0x4c, 0x0f, 0x0b, 0xe3, 0x9a, 0xd9, 0xce, 0x21, 0x73, 0xbc, 0xd9, 0xae, 0x15, 0x0b, 0x38, 0x93,
+	0x66, 0x7a, 0x58, 0x91, 0xd0, 0x42, 0x59, 0x26, 0x66, 0x6a, 0x8a, 0x53, 0xb7, 0x1b, 0xdb, 0x1b,
+	0x38, 0x2f, 0x79, 0x2b, 0xb4, 0x2e, 0x59, 0xeb, 0x28, 0xef, 0x8e, 0xea, 0x0e, 0xdb, 0x14, 0x27,
+	0xbf, 0xec, 0x0c, 0x86, 0x65, 0x0c, 0x3c, 0xff, 0x3e, 0x80, 0xd1, 0xd3, 0x9f, 0x11, 0xde, 0xc3,
+	0xb1, 0xf0, 0x22, 0x56, 0xc1, 0x7f, 0x31, 0x5e, 0xcc, 0xba, 0xfe, 0xad, 0x69, 0x9e, 0x86, 0xad,
+	0xa0, 0x1d, 0xe0, 0x33, 0x5c, 0x6e, 0x58, 0x93, 0xee, 0x47, 0x52, 0xc2, 0xeb, 0x64, 0x7d, 0xd7,
+	0x2d, 0x1f, 0xa7, 0xcf, 0x31, 0x70, 0xcb, 0xaa, 0xfa, 0x58, 0xff, 0x97, 0xf3, 0xfd, 0x78, 0x3b,
+	0xc0, 0x5b, 0x38, 0x12, 0x7e, 0x15, 0xf2, 0xa1, 0xef, 0x7b, 0x5f, 0xfd, 0x5b, 0xd6, 0x5c, 0xea,
+	0xee, 0x27, 0x00, 0x00, 0xff, 0xff, 0x5d, 0x71, 0x43, 0xa8, 0xbe, 0x01, 0x00, 0x00,
 }
 
 // Reference imports to suppress errors if they are not otherwise used.
@@ -214,6 +214,7 @@ const _ = grpc.SupportPackageIsVersion4
 type HandwritingClient interface {
 	Recognize(ctx context.Context, in *Matrix, opts ...grpc.CallOption) (*Result, error)
 	SetNeuralNetworkData(ctx context.Context, in *NeuralNetworkRaw, opts ...grpc.CallOption) (*None, error)
+	GetNeuralNetworkData(ctx context.Context, in *None, opts ...grpc.CallOption) (*NeuralNetworkRaw, error)
 	ReTrain(ctx context.Context, in *None, opts ...grpc.CallOption) (*None, error)
 }
 
@@ -243,6 +244,15 @@ func (c *handwritingClient) SetNeuralNetworkData(ctx context.Context, in *Neural
 	return out, nil
 }
 
+func (c *handwritingClient) GetNeuralNetworkData(ctx context.Context, in *None, opts ...grpc.CallOption) (*NeuralNetworkRaw, error) {
+	out := new(NeuralNetworkRaw)
+	err := c.cc.Invoke(ctx, "/handwriting.Handwriting/getNeuralNetworkData", in, out, opts...)
+	if err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
 func (c *handwritingClient) ReTrain(ctx context.Context, in *None, opts ...grpc.CallOption) (*None, error) {
 	out := new(None)
 	err := c.cc.Invoke(ctx, "/handwriting.Handwriting/reTrain", in, out, opts...)
@@ -256,6 +266,7 @@ func (c *handwritingClient) ReTrain(ctx context.Context, in *None, opts ...grpc.
 type HandwritingServer interface {
 	Recognize(context.Context, *Matrix) (*Result, error)
 	SetNeuralNetworkData(context.Context, *NeuralNetworkRaw) (*None, error)
+	GetNeuralNetworkData(context.Context, *None) (*NeuralNetworkRaw, error)
 	ReTrain(context.Context, *None) (*None, error)
 }
 
@@ -269,6 +280,9 @@ func (*UnimplementedHandwritingServer) Recognize(ctx context.Context, req *Matri
 func (*UnimplementedHandwritingServer) SetNeuralNetworkData(ctx context.Context, req *NeuralNetworkRaw) (*None, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method SetNeuralNetworkData not implemented")
 }
+func (*UnimplementedHandwritingServer) GetNeuralNetworkData(ctx context.Context, req *None) (*NeuralNetworkRaw, error) {
+	return nil, status.Errorf(codes.Unimplemented, "method GetNeuralNetworkData not implemented")
+}
 func (*UnimplementedHandwritingServer) ReTrain(ctx context.Context, req *None) (*None, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method ReTrain not implemented")
 }
@@ -313,6 +327,24 @@ func _Handwriting_SetNeuralNetworkData_Handler(srv interface{}, ctx context.Cont
 	return interceptor(ctx, in, info, handler)
 }
 
+func _Handwriting_GetNeuralNetworkData_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
+	in := new(None)
+	if err := dec(in); err != nil {
+		return nil, err
+	}
+	if interceptor == nil {
+		return srv.(HandwritingServer).GetNeuralNetworkData(ctx, in)
+	}
+	info := &grpc.UnaryServerInfo{
+		Server:     srv,
+		FullMethod: "/handwriting.Handwriting/GetNeuralNetworkData",
+	}
+	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
+		return srv.(HandwritingServer).GetNeuralNetworkData(ctx, req.(*None))
+	}
+	return interceptor(ctx, in, info, handler)
+}
+
 func _Handwriting_ReTrain_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
 	in := new(None)
 	if err := dec(in); err != nil {
@@ -343,6 +375,10 @@ var _Handwriting_serviceDesc = grpc.ServiceDesc{
 			MethodName: "setNeuralNetworkData",
 			Handler:    _Handwriting_SetNeuralNetworkData_Handler,
 		},
+		{
+			MethodName: "getNeuralNetworkData",
+			Handler:    _Handwriting_GetNeuralNetworkData_Handler,
+		},
 		{
 			MethodName: "reTrain",
 			Handler:    _Handwriting_ReTrain_Handler,

handwriting/handwritingui/handwritingengine.cpp

@@ -51,6 +51,7 @@ static const QtProtobuf::DoubleList emptyData{0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.
 
 HandwritingEngine::HandwritingEngine(QObject *parent) : QObject(parent)
   , m_client(new handwriting::HandwritingClient)
+  , m_result(0)
 {
 
     auto chan = std::shared_ptr<QtProtobuf::QGrpcHttp2Channel>(new QtProtobuf::QGrpcHttp2Channel(QUrl("http://localhost:65001"), QtProtobuf::InsecureCredentials()|NoneCredencials()));
@@ -65,8 +66,8 @@ void HandwritingEngine::retrain() {
 
 void HandwritingEngine::recognize()
 {
-    m_client->recognize(m_matrix, this, [](QtProtobuf::QGrpcAsyncReply *reply) {
-        qDebug() << "Recognition result" << reply->read<handwriting::Result>().resultCharacter();
+    m_client->recognize(m_matrix, this, [this](QtProtobuf::QGrpcAsyncReply *reply) {
+        setResult(reply->read<handwriting::Result>().resultCharacter());
     });
 }
 

handwriting/handwritingui/main.qml

@@ -63,8 +63,8 @@ ApplicationWindow {
                         var diffX = mouse.x - centerX
                         var diffY = mouse.y - centerY
                         var dense = Math.sqrt(diffX * diffX + diffY * diffY)
-                        if (dense < tileSize) {
-                            var newColor = color.r + (tileSize - dense)/tileSize
+                        if (dense < tileSize*1.5) {
+                            var newColor = color.r + (tileSize*1.5 - dense)/tileSize*1.5
                             var newValue = newColor - 0.2
                             if (newColor > 1.0) {
                                 newColor = 1.0
@@ -89,6 +89,7 @@ ApplicationWindow {
     }
 
     Column {
+        id: controls
         anchors.left: drawingArea.right
         anchors.top: parent.top
         anchors.margins: 10
@@ -140,6 +141,16 @@ ApplicationWindow {
         }
     }
 
+    Text {
+        id: result
+        anchors.top: controls.bottom
+        anchors.topMargin: 100
+        anchors.horizontalCenter: controls.horizontalCenter
+        color: "#003b6f" //Tardis blue :)
+        font.pixelSize: 40
+        text: hwengine.result
+    }
+
     FileDialog {
         id: neuralNetworkImport
         selectMultiple: false

neuralnetwork/neuralnetwork/neuralnetwork.go

@@ -240,9 +240,9 @@ func (nn *NeuralNetwork) TrainOnline(trainer training.Trainer, epocs int) {
 }
 
 func (nn *NeuralNetwork) TrainBatch(trainer training.Trainer, epocs int) {
-	fmt.Printf("Start training in %v threads\n", 2*runtime.NumCPU())
+	fmt.Printf("Start training in %v threads\n", runtime.NumCPU())
 	for t := 0; t < epocs; t++ {
-		batchWorkers := nn.runBatchWorkers(2*runtime.NumCPU(), trainer)
+		batchWorkers := nn.runBatchWorkers(runtime.NumCPU(), trainer)
 
 		for l := 1; l < nn.LayerCount; l++ {
 			bGradient, ok := nn.BGradient[l].(BatchGradientDescent)