diff --git a/examples/SuperResolution/1090_usuperesolutionapp21test.pas b/examples/SuperResolution/1090_usuperesolutionapp21test.pas
new file mode 100644
index 00000000..03533ccb
--- /dev/null
+++ b/examples/SuperResolution/1090_usuperesolutionapp21test.pas
@@ -0,0 +1,510 @@
+unit usuperesolutionapp_mX4;
+//{$mode objfpc}{$H+}
+
+interface
+
+{uses
+  Classes, SysUtils, FileUtil, Forms, Controls, Graphics, Dialogs, StdCtrls,
+  ExtCtrls, neuraldatasets, neuraldatasetsv, fpimage, IntfGraphics, lcltype, Buttons,
+  neuralnetwork, neuralvolume, neuralvolumev;   }
+
+type
+  { TForm1 }
+  TForm1 = {class(}TForm;
+  var
+    BitBtn13: TBitBtn;
+    CheckPause: TCheckBox;
+    Image1: TImage;
+    ImageHE: TImage;
+    ImageVE: TImage;
+    ImageGray: TImage;
+    Label1: TLabel;
+    Label2: TLabel;
+    Label3: TLabel;
+    Label4: TLabel;
+    Label5: TLabel;
+    procedure TForm1BitBtn13Click(Sender: TObject);
+    procedure TForm1FormClose(Sender: TObject; var CloseAction: TCloseAction);
+    procedure TForm1FormCreate(Sender: TObject);
+  //private
+  var
+    FWantQuit: boolean;
+    function CheckBinFiles: boolean;
+  //public
+
+  //end;
+  
+ function CreateResizingNN(SizeX, SizeY: integer; FileName: string): THistoricalNets;
+ procedure LoadResizingWeights(NN:TNNet; FileName:string; FailIfNotFound:boolean{ = false});
+ 
+
+  { TEvolutionary }
+var
+  Form1: TForm1;
+
+implementation
+//uses usuperresolutionexample;
+
+ //unit usuperresolutionexample;
+// need PReModel_NN_Res = '.\examples\super-resolution-7-64-sep.nn';
+
+const
+  csExampleBaseFileName = '1090_super-resolution-7-64-sep';
+  csExampleFileName {:string} = '1090_super-resolution-7-64-sep.nn';
+  csExampleBottleNeck {:integer =} = 16;
+  csExampleNeuronCount = 64;
+  csExampleLayerCount = 7;
+  csExampleIsSeparable {:boolean} =  true;          
+  
+function CreateResizingNN(SizeX, SizeY: integer; FileName: string): THistoricalNets;
+//var
+ // NN: THistoricalNets;
+begin
+  result := THistoricalNets.Create();
+  result.AddSuperResolution({pSizeX=}SizeX, {pSizeY=}SizeY,
+    {BottleNeck=}csExampleBottleNeck, {pNeurons=}csExampleNeuronCount,
+    {pLayerCnt=}csExampleLayerCount, {IsSeparable=}csExampleIsSeparable);
+  LoadResizingWeights(result, FileName, true);
+  //Result := NN;
+  //nn.free;
+end;
+
+procedure LoadResizingWeights(NN:TNNet; FileName:string; FailIfNotFound:boolean {= false});
+begin
+  if FileExists(FileName) then begin
+    NN.LoadDataFromFile(FileName);
+    WriteLn('Neural network file found:'+ FileName);
+  end else
+  if FileExists('./model/'+FileName) then begin
+    NN.LoadDataFromFile('./model/'+FileName);
+    WriteLn('Neural network file found at ./model : '+ FileName);
+  end else begin
+    if FailIfNotFound then begin
+      WriteLn('ERROR: '+FileName+'can''t be found. Please run SuperResolutionTrain.');
+      //ReadLn();
+    end;
+  end;
+end;
+
+var csTinyImageLabel: array[0..9] of string;
+procedure setClassifierLabelsTiny;
+begin
+  csTinyImageLabel[0]:= 'airplane';
+  csTinyImageLabel[1]:= 'automobile';
+  csTinyImageLabel[2]:= 'bird';
+  csTinyImageLabel[3]:= 'cat';
+  csTinyImageLabel[4]:= 'deer';
+  csTinyImageLabel[5]:= 'dog';
+  csTinyImageLabel[6]:= 'frog';
+  csTinyImageLabel[7]:= 'horse';
+  csTinyImageLabel[8]:= 'ship';
+  csTinyImageLabel[9]:= 'truck';
+end;  
+
+ const TBATCH= 'C:\Program Files\Streaming\maxbox4\maxbox47590\maxbox4\cifar-10-batches-bin\test_batch.bin';
+
+procedure TestBatchPerformance(aNN: TNNet; Testpath: string; aimgvolumes:TNNetVolumeList);
+var Rate,Loss,ErrorSum,LastError: TNeuralFloat;
+    ImgVolumes: TNNetVolumeList;
+begin
+    rate:=0; loss:=0;
+    ErrorSum:=0;
+    ImgVolumes:= TNNetVolumeList.Create(true);
+    try 
+    // creates required volumes to store images
+      for it:= 0 to 9999 do 
+        ImgVolumes.Add(TNNetVolume.Create());
+      loadCifar10Dataset(ImgVolumes, TBATCH, 0,csEncodeRGB); //"test_batch.bin"  }
+      writeln('testpath '+testpath)
+      //if pos('Resize40',Testpath) > 0 then
+         ImgVolumes.ResizeImage(64, 64); 
+      WriteLn(' Testbatch size: '+ itoa(ImgVolumes.gettotalsize));
+      WriteLn(' dbug volumescount: '+ itoa(ImgVolumes.Count));
+      //ImgVolumes[it].ReSize(40,40,3);
+      TestBatch(aNN, ImgVolumes, 500, rate,loss,ErrorSum);
+      writeln('Testbatch score: '+Format('Rate:%.4f, Loss:%.4f, ErrorSum:%.4f',
+                                                      [rate, loss, ErrorSum]));
+      label2.Caption:= format('Score Ø %.2f%% ',[rate*100]);
+    finally
+      ImgVolumes.Free;
+  end;  
+end;     
+
+
+//{$R *.lfm}
+
+procedure loadSuperForm;
+begin
+Form1:= TForm1.create(self)
+with Form1 do begin
+  Left := 587
+  Height:= 491
+  Top:= 249
+  Width:= 958
+  BorderStyle:= bsSingle; //bsDialog
+  Caption:= 'CIFAR-10 Super Resolution App maXbox4'
+  ClientHeight:= 491
+  ClientWidth:= 958
+  icon.loadfromresourcename(hinstance,'ZCUBE');
+  //DesignTimePPI:= 120
+  OnClose:= @TForm1FormClose;
+  OnCreate:= @TForm1FormCreate;
+  //LCLVersion:= '2.0.2.0'
+  Image1:= TImage.create(form1)
+  Show;
+  with image1 do begin
+    parent:= form1
+    Left:= 40
+    Height:= 32
+    Top:= 50
+    Width:= 32
+    Stretch:= True
+  end;
+  Label1:= TLabel.create(form1)
+  with label1 do begin
+   parent:= form1
+    Left:= 580
+    Height:= 20
+    Top:= 344
+    Width:= 73
+    Caption:= 'Label Class'
+    font.size:= 14;
+    ParentColor:= False
+    ParentFont:= False
+  end ;
+  ImageGray:= TImage.create(form1)
+  with imagegray do begin
+   parent:= form1;
+    Left:= 220
+    Height:= 64
+    Top:= 50
+    Width:= 64
+    Stretch:= True
+  end;
+  //&object ImageHE: TImage
+  Imagehe:= TImage.create(form1)
+  with imagehe do begin
+   parent:= form1;
+    Left:= 400
+    Height:= 128
+    Top:= 50
+    Width:= 128
+    Stretch:= True
+  end;
+  Imageve:= TImage.create(form1)
+  with imageve do begin
+   parent:= form1;
+    Left:= 580
+    Height:= 256
+    Top:= 50
+    Width:= 256
+    Stretch:= True
+  end ;
+   BitBtn13:=  TBitBtn.create(form1)
+   with bitbtn13 do begin
+    parent:= form1
+    Left:= 40
+    Height:= 45
+    Top:= 344
+    Width:= 182
+    Caption:= ' RUN SUPER'
+    OnClick:= @Tform1BitBtn13Click;
+    glyph.loadfromresourcename(hinstance,'WET2');
+    ParentFont:= False
+    TabOrder:= 0
+  end ;
+  Label2:= TLabel.create(form1)
+  with label2 do begin
+   parent:= form1
+    Left:= 40
+    Height:= 20
+    Top:= 25
+    Width:= 96
+    Caption:= 'Original 32x32'
+    ParentColor:= False
+    ParentFont:= False
+  end;
+  Label3:= TLabel.create(form1)
+  with label3 do begin
+   parent:= form1
+    Left:= 220
+    Height:= 20
+    Top:= 25
+    Width:= 113
+    Caption:= 'Computed 64x64'
+    ParentColor:= False
+    ParentFont:= False
+  end ;
+  Label4:= TLabel.create(form1)
+  with label4 do begin
+   parent:= form1
+    Left:= 400
+    Height:= 20
+    Top:= 25
+    Width:= 129
+    Caption:= 'Computed 128x128'
+    ParentColor:= False
+    ParentFont:= False
+  end ;
+  Label5:= TLabel.create(form1)
+  with label5 do begin
+   parent:= form1
+    Left:= 580
+    Height:= 20
+    Top:= 25
+    Width:= 129
+    Caption:= 'Computed 256x256'
+    ParentColor:= False
+    ParentFont:= False
+  end;
+   CheckPause:= TCheckBox.create(form1)
+   with checkpause do begin
+    parent:= form1
+    Left:= 40
+    Height:= 24
+    Top:= 400
+    Width:= 191
+    Caption:= 'Run Pause'
+    ParentFont:= False
+    TabOrder:= 1
+  end ;
+ end;
+ TForm1FormCreate(self);
+end; 
+
+{ TForm1 }
+procedure TForm1BitBtn13Click(Sender: TObject);
+var
+  NN2, NN3, NN41: THistoricalNets;
+  Img: TTinyImage;
+  //cifarFile: TTInyImageFile;
+  I, K: integer;
+  ImgVolumes: TNNetVolumeList;
+  Volume: TNNetVolume;
+  vInput, vOutput, vBigOutput, vDisplay: TNNetVolume;
+begin
+  if not(CheckBinFiles()) then begin
+    writeln('CheckBinFiles not OK');
+    //exit;
+  end;
+
+  Randomize;
+  BitBtn13.Enabled := false;
+  writeln('Creating Neural Network...');
+  ImgVolumes:= TNNetVolumeList.Create(true);
+  
+  writeln('Loading Tiny Cifar Images...');
+     for I:= 0 to 9999 do begin
+       Volume := TNNetVolume.Create();
+       ImgVolumes.Add(Volume);
+     end;
+   loadCifar10Dataset6(ImgVolumes, 1, 0,csEncodeRGB);   //1 is data_batch_1.bin
+   writeln('tot size1 '+itoa(ImgVolumes.GetTotalSize( ))+ ' '+
+                                                 itoa(imgvolumes.count));    //*)
+   //TranslateCifar10VolumesToMachineAnimal(ImgVolumes);                                              
+                                                 
+   //CreateCifar10Volumes(ImgVolumes,ImgVolumes, ImgVolumes, csencodergb);                                              
+    for I:= 2000 to 9999 do begin
+        ImgVolumes.items[i].resize(0,0,0);  
+        ImgVolumes.items[i].free; 
+        //ImgVolumes.items[i]:= Nil; 
+     end;    
+    writeln('tot size2 '+itoa(ImgVolumes.GetTotalSize( ))+ ' '+
+                                                 itoa(imgvolumes.count));
+
+ //try
+     NN2 := CreateResizingNN(32, 32, csExampleFileName);
+     //NN2.DebugStructure();
+     writeln('after debug struct');
+     sleep(200);
+     //TestBatchPerformance(NN2, csExampleFileName);             
+     NN3 := CreateResizingNN(64, 64, csExampleFileName);
+     sleep(300)
+     try
+       NN41 := CreateResizingNN(128, 128, csExampleFileName);
+     except
+       writeln('mem err in create resize..');
+     end;  
+     //NN2 := CreateResizingNN(32, 32, csExampleFileName);
+       
+   try  
+   //AssignFile(cifarFile, 'data_batch_1.bin');
+   //for I:= 100 to 5999 do 
+     //   ImgVolumes.delete(imgvolumes[i]);
+    
+                                               
+   //Reset(cifarFile);
+   I := 1;
+   writeln('image count and testbatch: '+itoa(ImgVolumes.count ));
+  // TestBatchPerformance(NN3, csExampleFileName, imgvolumes);  
+  //while not EOF(cifarFile) do
+  for it:= 1 to 200 do begin
+    //Read(cifarFile, Img);
+    LoadNNetVolumeIntoTinyImage4(imgvolumes[it], img);
+    Label1.Caption := 'label: '+csTinyImageLabel[Img.bLabel];
+    writeln('debug '+csTinyImageLabel[Img.bLabel]);
+    Volume := TNNetVolume.Create();
+    writeln('debug1 '+csTinyImageLabel[Img.bLabel]);
+    LoadTinyImageIntoNNetVolume1(Img, Volume);
+    //imgvolumes[it].divi75(64);
+    Volume.Divi75(64);
+    Volume.Sub19(2);
+    //imgvolumes[it].Sub64(2);
+    //writeln('debug3 '+csTinyImageLabel[Img.bLabel]);
+    ImgVolumes.Add(Volume);
+    
+    if (I mod 100 = 0) then begin
+      LoadTinyImageIntoTImage(Img, Image1);
+      LoadTinyImageIntoTImage(Img, ImageGray);
+      LoadTinyImageIntoTImage(Img, ImageHE);
+      LoadTinyImageIntoTImage(Img, ImageVE);
+      Image1.Width := 128;
+      Image1.Height := 128;
+      ImageGray.Width := 128;
+      ImageGray.Height := 128;
+      ImageHE.Width := 128;
+      ImageHE.Height := 128;
+      ImageVE.Width := 256;
+      ImageVE.Height := 256;
+      Application.ProcessMessages;
+    end;
+    inc(I);
+  end;
+  //CloseFile(cifarFile);
+
+  vInput     := TNNetVolume.Create0(32,32,3,0);
+  vOutput    := TNNetVolume.Create0(64,64,3,0);
+  vDisplay   := TNNetVolume.Create();
+  vBigOutput := TNNetVolume.Create0(64,64,3,0);
+
+  WriteLn('128x128 -> 256x256 Neural network has: ');
+  WriteLn(' Layers: '+itoa( NN41.CountLayers()  ));
+  WriteLn(' Neurons:'+itoa(NN41.CountNeurons() ));
+  WriteLn(' Weights:'+itoa(NN41.CountWeights() ));
+  WriteLn('Computing SuperResolution...');
+
+  while not (FWantQuit) do begin
+    I := random(1000);
+    vInput.Copy(ImgVolumes[I]);
+    Label1.Caption:= 'label: '+csTinyImageLabel[ ImgVolumes[I].Tag ];
+    vDisplay.Copy(vInput);
+    vDisplay.Add13(2);
+    vDisplay.Mul26(64);
+    LoadNNetVolumeIntoTinyImage4(vDisplay, Img);
+    //writeln('debug4 '+csTinyImageLabel[Img.bLabel]);
+    LoadTinyImageIntoTImage(Img, Image1);
+    
+    NN2.Compute65(vInput,0); NN2.GetOutput(vBigOutput);
+    vDisplay.Copy(vBigOutput);
+    vDisplay.Add13(2);
+    vDisplay.Mul26(64);
+    //LoadVolumeIntoTImage(vDisplay, ImageVE, csEncodeRGB);
+    LoadVolumeIntoImage(vDisplay, ImageGray);
+    ImageGray.Width := 128;
+    ImageGray.Height := 128;
+
+    NN3.Compute65(vBigOutput,0);  NN3.GetOutput(vBigOutput);
+    vDisplay.Copy(vBigOutput);
+    vDisplay.Add13(2);
+    vDisplay.Mul26(64);
+    LoadVolumeIntoImage(vDisplay, ImageHE);
+    ImageHE.Width := 128;
+    ImageHE.Height := 128;
+
+    //vbigoutput.resize(128,128,3)
+    NN41.Compute65(vBigOutput,0);  NN41.GetOutput(vBigOutput);
+    //NN3.Compute65(vBigOutput,0);  NN3.GetOutput(vBigOutput);
+    vDisplay.Copy(vBigOutput);
+    vDisplay.Add13(2);
+    vDisplay.Mul26(64);   //64
+    LoadVolumeIntoImage(vDisplay, ImageVE);
+    ImageVE.Width := 256;
+    ImageVE.Height := 256;  //}
+
+    for K := 1 to 50 do begin                                 
+      Application.ProcessMessages();
+      Sleep(100);
+    end;
+    while CheckPause.Checked do begin
+      Application.ProcessMessages();
+      Sleep(100);
+      FWantQuit := true; 
+      CheckPause.Caption:= 'Run Pause or Close'
+    end;
+  end; //while
+  finally
+     NN2.Free;
+     NN3.Free;
+     NN41.Free; //}
+     vBigOutput.Free;
+     vDisplay.Free;
+     vInput.Free;
+     vOutput.Free;                             
+     ImgVolumes.Free;
+     BitBtn13.Enabled := true;
+     FWantQuit := false; 
+     writeln('debug finally and frees..');
+  end;   
+end;
+
+procedure TForm1FormClose(Sender: TObject; var CloseAction: TCloseAction);
+begin
+  FWantQuit := true;
+    {NN2.Free;
+     NN3.Free;
+     //NN4.Free;
+     vBigOutput.Free;
+     vDisplay.Free;
+     vInput.Free;
+     vOutput.Free;                             
+     ImgVolumes.Free;   }
+  writeln('form close...');
+end;
+
+procedure TForm1FormCreate(Sender: TObject);
+begin
+  FWantQuit := false;
+  writeln('form createp sueperres Iü');
+end;
+
+function CheckBinFiles: boolean;
+begin
+  Result := true;
+  if not (FileExists('data_batch_1.bin')) then begin
+    Result := false;
+    ShowMessage('CIFAR-10 files have not been found.' + Chr(13) +
+      'Please download from https://www.cs.toronto.edu/~kriz/cifar-10-binary.tar.gz');
+  end;
+end;
+
+  var
+  NN, NN3: THistoricalNets;
+
+begin //@main
+
+  setClassifierLabelsTiny;
+  loadsuperForm();
+  
+ (* NN:= THistoricalNets.Create();
+  NN.AddSuperResolution({pSizeX=}128, {pSizeY=}128,
+    {BottleNeck=}csExampleBottleNeck, {pNeurons=}csExampleNeuronCount,
+    {pLayerCnt=}csExampleLayerCount, {IsSeparable=}csExampleIsSeparable);
+  //LoadResizingWeights(NN, FileName, true);
+  LoadResizingWeights(NN, csExampleFileName, true);
+  writeln(objtostr(nn));
+  nn.free;  *)
+  
+  nn:= CreateResizingNN(128, 128, csExampleFileName);
+  writeln(objtostr(nn));
+  writeln('layers '+itoa(nn.CountLayers()));
+  nn.free;
+  
+  {NN3:= CreateResizingNN(64, 64, csExampleFileName);
+  writeln(objtostr(nn3));
+  nn3.free; }
+
+ end.
+End.
+
+ref: https://github.com/maxkleiner/neural-api/blob/master/examples/SuperResolution/usuperesolutionapp.pas
+