<![CDATA[Forums - VisionLab]]> http://mitov.com/forum/ Wed, 17 Jun 2026 00:00:16 +0000 MyBB <![CDATA[Haar Cascade XML, what format?]]> http://mitov.com/forum/thread-3529.html Tue, 20 Dec 2016 15:54:50 -0500 http://mitov.com/forum/thread-3529.html
I seems that the current xml does not match what was included with the lab packs.

This being the case, I can't use a new haar trainer to generate xml for my project, and I current;y have no way to determine what version/variant of an older haar trainer would be compatible.

Help!

Example:
LabPack
Code:
<opencv_storage>
<haarcascade_frontaleye type_id="opencv-haar-classifier">
  <size>
    20 20</size>
  <stages>
    <_>
      <!-- stage 0 -->
      <trees>
        <_>
          <!-- tree 0 -->
          <_>
            <!-- root node -->
            <feature>
              <rects>
                <_>
                  0 8 20 12 -1.</_>
                <_>
                  0 14 20 6 2.</_></rects>
              <tilted>0</tilted></feature>
            <threshold>0.1296395957469940</threshold>
            <left_val>-0.7730420827865601</left_val>
            <right_val>0.6835014820098877</right_val></_></_>
        <_>
          <!-- tree 1 -->
          <_>
            <!-- root node -->
            <feature>
              <rects>
                <_>
                  9 1 4 15 -1.</_>
                <_>
                  9 6 4 5 3.</_></rects>
              <tilted>0</tilted></feature>
            <threshold>-0.0463268086314201</threshold>
            <left_val>0.5735275149345398</left_val>
            <right_val>-0.4909768998622894</right_val></_></_>
        <_>

Current OpenCV
Code:
<opencv_storage>
<cascade type_id="opencv-cascade-classifier"><stageType>BOOST</stageType>
  <featureType>HAAR</featureType>
  <height>20</height>
  <width>20</width>
  <stageParams>
    <maxWeakCount>93</maxWeakCount></stageParams>
  <featureParams>
    <maxCatCount>0</maxCatCount></featureParams>
  <stageNum>24</stageNum>
  <stages>
    <_>
      <maxWeakCount>6</maxWeakCount>
      <stageThreshold>-1.4562760591506958e+00</stageThreshold>
      <weakClassifiers>
        <_>
          <internalNodes>
            0 -1 0 1.2963959574699402e-01</internalNodes>
          <leafValues>
            -7.7304208278656006e-01 6.8350148200988770e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 1 -4.6326808631420135e-02</internalNodes>
          <leafValues>
            5.7352751493453979e-01 -4.9097689986228943e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 2 -1.6173090785741806e-02</internalNodes>
          <leafValues>
            6.0254341363906860e-01 -3.1610709428787231e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 3 -4.5828841626644135e-02</internalNodes>
          <leafValues>
            6.4177548885345459e-01 -1.5545040369033813e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 4 -5.3759619593620300e-02</internalNodes>
          <leafValues>
            5.4219317436218262e-01 -2.0480829477310181e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 5 3.4171190112829208e-02</internalNodes>
          <leafValues>
            -2.3388190567493439e-01 4.8410901427268982e-01</leafValues></_></weakClassifiers></_>
    <_>
]]>
I seems that the current xml does not match what was included with the lab packs.

This being the case, I can't use a new haar trainer to generate xml for my project, and I current;y have no way to determine what version/variant of an older haar trainer would be compatible.

Help!

Example:
LabPack
Code:
<opencv_storage>
<haarcascade_frontaleye type_id="opencv-haar-classifier">
  <size>
    20 20</size>
  <stages>
    <_>
      <!-- stage 0 -->
      <trees>
        <_>
          <!-- tree 0 -->
          <_>
            <!-- root node -->
            <feature>
              <rects>
                <_>
                  0 8 20 12 -1.</_>
                <_>
                  0 14 20 6 2.</_></rects>
              <tilted>0</tilted></feature>
            <threshold>0.1296395957469940</threshold>
            <left_val>-0.7730420827865601</left_val>
            <right_val>0.6835014820098877</right_val></_></_>
        <_>
          <!-- tree 1 -->
          <_>
            <!-- root node -->
            <feature>
              <rects>
                <_>
                  9 1 4 15 -1.</_>
                <_>
                  9 6 4 5 3.</_></rects>
              <tilted>0</tilted></feature>
            <threshold>-0.0463268086314201</threshold>
            <left_val>0.5735275149345398</left_val>
            <right_val>-0.4909768998622894</right_val></_></_>
        <_>

Current OpenCV
Code:
<opencv_storage>
<cascade type_id="opencv-cascade-classifier"><stageType>BOOST</stageType>
  <featureType>HAAR</featureType>
  <height>20</height>
  <width>20</width>
  <stageParams>
    <maxWeakCount>93</maxWeakCount></stageParams>
  <featureParams>
    <maxCatCount>0</maxCatCount></featureParams>
  <stageNum>24</stageNum>
  <stages>
    <_>
      <maxWeakCount>6</maxWeakCount>
      <stageThreshold>-1.4562760591506958e+00</stageThreshold>
      <weakClassifiers>
        <_>
          <internalNodes>
            0 -1 0 1.2963959574699402e-01</internalNodes>
          <leafValues>
            -7.7304208278656006e-01 6.8350148200988770e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 1 -4.6326808631420135e-02</internalNodes>
          <leafValues>
            5.7352751493453979e-01 -4.9097689986228943e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 2 -1.6173090785741806e-02</internalNodes>
          <leafValues>
            6.0254341363906860e-01 -3.1610709428787231e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 3 -4.5828841626644135e-02</internalNodes>
          <leafValues>
            6.4177548885345459e-01 -1.5545040369033813e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 4 -5.3759619593620300e-02</internalNodes>
          <leafValues>
            5.4219317436218262e-01 -2.0480829477310181e-01</leafValues></_>
        <_>
          <internalNodes>
            0 -1 5 3.4171190112829208e-02</internalNodes>
          <leafValues>
            -2.3388190567493439e-01 4.8410901427268982e-01</leafValues></_></weakClassifiers></_>
    <_>
]]> <![CDATA[TVLHaarObjectDetect 4-5 Second Delay]]> http://mitov.com/forum/thread-3516.html Tue, 13 Dec 2016 13:48:45 -0500 http://mitov.com/forum/thread-3516.html The VLHaarDetectObjects control is updating a second VLImageDisplay.
The detect is working, but there seems to be 4-5 seconds behind the camera.

Any thoughts on optimizations, compression algorithms, etc.?

I tried recording the video, and using an AVI player as in the demo, but that also trails by 4-5 seconds.

If I use a single AVI player as in the demo, the delay isn't visible, but if I play the same source through two different AVI Players, the feed from the HaarDetect quickly gets (and stays) 4-5 seconds behind...]]> The VLHaarDetectObjects control is updating a second VLImageDisplay.
The detect is working, but there seems to be 4-5 seconds behind the camera.

Any thoughts on optimizations, compression algorithms, etc.?

I tried recording the video, and using an AVI player as in the demo, but that also trails by 4-5 seconds.

If I use a single AVI player as in the demo, the delay isn't visible, but if I play the same source through two different AVI Players, the feed from the HaarDetect quickly gets (and stays) 4-5 seconds behind...]]> <![CDATA[VLAxisCapture memory leak]]> http://mitov.com/forum/thread-3018.html Mon, 30 Sep 2013 11:19:25 -0400 http://mitov.com/forum/thread-3018.html
Might be in the Axis active X control, no way for me to know for sure (no source).

However I replaced the video source with a generic image generator (playing 9 frames from bitmap files on a loop) and the demo app memory use is stable so this doesn't appear to be related to motion detection or vldraw/layers/etc.]]>
Might be in the Axis active X control, no way for me to know for sure (no source).

However I replaced the video source with a generic image generator (playing 9 frames from bitmap files on a loop) and the demo app memory use is stable so this doesn't appear to be related to motion detection or vldraw/layers/etc.]]> <![CDATA[Cascade Classification for TVLHaarDetectObject]]> http://mitov.com/forum/thread-3001.html Mon, 08 Jul 2013 08:21:52 -0400 http://mitov.com/forum/thread-3001.html
TVLHaarDetectObject demands xml-file Cascade Classification.
How to create it?
What component allows to make it?
Whether there are examples?]]>
TVLHaarDetectObject demands xml-file Cascade Classification.
How to create it?
What component allows to make it?
Whether there are examples?]]> <![CDATA[Video tracker]]> http://mitov.com/forum/thread-932.html Mon, 15 Oct 2012 04:26:45 -0400 http://mitov.com/forum/thread-932.html I will like to use the VisionLab VCL components to
develop a real time video tracker.
I need to user to click on the video image and then use visionlab
to find a target to lock on.
Then I need to move a pan and tilt video camera according to
the target movement.
You can look to an example here:
http://www.trackertechnology.com/demos.html

Is there a way of doing so, and if so what components of visionlab
should I use and in what way.
If you can send me an example I will appreciate it.
I looked at the TargetsTrackDemo.exe example but I need to keep tracking the object near the mouse click all the time. In the example
the acquired targets change all the time.
Thanks Allot
Raz]]> I will like to use the VisionLab VCL components to
develop a real time video tracker.
I need to user to click on the video image and then use visionlab
to find a target to lock on.
Then I need to move a pan and tilt video camera according to
the target movement.
You can look to an example here:
http://www.trackertechnology.com/demos.html

Is there a way of doing so, and if so what components of visionlab
should I use and in what way.
If you can send me an example I will appreciate it.
I looked at the TargetsTrackDemo.exe example but I need to keep tracking the object near the mouse click all the time. In the example
the acquired targets change all the time.
Thanks Allot
Raz]]> <![CDATA[TVLHaarDetectObject]]> http://mitov.com/forum/thread-355.html Mon, 21 May 2012 01:42:19 -0400 http://mitov.com/forum/thread-355.html I downloaded Face Detection Demo and try to rewrite pascal into c++ builder.
In OnDetectedObjects event handler in c++ code, I wrote below.

// code starts
void __fastcall TForm1::VLHaarDetectObjects1DetectedObjects(TObject *Sender, TVLDetectedObjects *Objects)

{
int i,j;
TVLImageRect ARectangle;
TVLBitmap* ABitmap;


CodeSite->Send("Objects->Count:"+IntToStr(Objects->Count));
ABitmap = new TVLBitmap();
VLHaarDetectObjects1->CurrentImage->ToBitmap(ABitmap);

ABitmap->Canvas->Font->Color = clWhite;
for(i=0;i<Objects->Count;i++){
ARectangle = Objects->Items[i]->Rectangle;
ABitmap->Canvas->Brush->Style = bsClear;
ABitmap->Canvas->Pen->Color = clRed;
ABitmap->Canvas->Ellipse(ARectangle.X,ARectangle.Y,ARectangle.X+ARectangle.Width,ARectangle.Y+​ARectangle.Height);
ABitmap->Canvas->TextOutW(ARectangle.X+ARectangle.Width/2,ARectangle.Y+ARectangle.Height/2,Objects->Items[i]->Name);

ABitmap->Canvas->Pen->Color = clYellow;
for(j=0;j<Objects->Items[i]->Objects->Count;j++){
ARectangle = Objects->Items[i]->Objects->Items[j]->Rectangle;
ABitmap->Canvas->Ellipse(ARectangle.X,ARectangle.Y,ARectangle.X+ARectangle.Width,ARectangle.Y+​ARectangle.Height);
ABitmap->Canvas->TextOutW(ARectangle.X+ARectangle.Width/2,ARectangle.Y+ARectangle.Height/2,
Objects->Items[i]->Objects->Items[j]->Name);
}
}

VLImageDisplay2->DisplayBitmap(ABitmap);
delete ABitmap;

}
// code ends.

It seems the event handler is called, but "Objects->Count" is always equal to zero. Objects->Count should have a value more than 0.
Suggestions and advice are appreciated.
Thanks in advance.

Compiler: RadStudio XE2(c++ Builder) update4
Component ver: 5.02]]> I downloaded Face Detection Demo and try to rewrite pascal into c++ builder.
In OnDetectedObjects event handler in c++ code, I wrote below.

// code starts
void __fastcall TForm1::VLHaarDetectObjects1DetectedObjects(TObject *Sender, TVLDetectedObjects *Objects)

{
int i,j;
TVLImageRect ARectangle;
TVLBitmap* ABitmap;


CodeSite->Send("Objects->Count:"+IntToStr(Objects->Count));
ABitmap = new TVLBitmap();
VLHaarDetectObjects1->CurrentImage->ToBitmap(ABitmap);

ABitmap->Canvas->Font->Color = clWhite;
for(i=0;i<Objects->Count;i++){
ARectangle = Objects->Items[i]->Rectangle;
ABitmap->Canvas->Brush->Style = bsClear;
ABitmap->Canvas->Pen->Color = clRed;
ABitmap->Canvas->Ellipse(ARectangle.X,ARectangle.Y,ARectangle.X+ARectangle.Width,ARectangle.Y+​ARectangle.Height);
ABitmap->Canvas->TextOutW(ARectangle.X+ARectangle.Width/2,ARectangle.Y+ARectangle.Height/2,Objects->Items[i]->Name);

ABitmap->Canvas->Pen->Color = clYellow;
for(j=0;j<Objects->Items[i]->Objects->Count;j++){
ARectangle = Objects->Items[i]->Objects->Items[j]->Rectangle;
ABitmap->Canvas->Ellipse(ARectangle.X,ARectangle.Y,ARectangle.X+ARectangle.Width,ARectangle.Y+​ARectangle.Height);
ABitmap->Canvas->TextOutW(ARectangle.X+ARectangle.Width/2,ARectangle.Y+ARectangle.Height/2,
Objects->Items[i]->Objects->Items[j]->Name);
}
}

VLImageDisplay2->DisplayBitmap(ABitmap);
delete ABitmap;

}
// code ends.

It seems the event handler is called, but "Objects->Count" is always equal to zero. Objects->Count should have a value more than 0.
Suggestions and advice are appreciated.
Thanks in advance.

Compiler: RadStudio XE2(c++ Builder) update4
Component ver: 5.02]]> <![CDATA[Simple Template Maching]]> http://mitov.com/forum/thread-100.html Tue, 31 Jan 2012 03:45:06 -0500 http://mitov.com/forum/thread-100.html How can I implement an image template matching algorithm.?
Does anyone have some exsample code ? all any suggesion?

Thanks in advance]]> How can I implement an image template matching algorithm.?
Does anyone have some exsample code ? all any suggesion?

Thanks in advance]]> <![CDATA[Freezing]]> http://mitov.com/forum/thread-68.html Wed, 07 Dec 2011 07:10:46 -0500 http://mitov.com/forum/thread-68.html
I am trying use VLDSCapture to capture my webcam video and i linked the output pin to VLHaarDetectObjects (it is configured like the example of facedetect to try detect my face) but when i start the VLHaarDetectObjects the VLImageDisplay is freezing. If i stop the VLHaarDetectObjects, my VLImageDisplay becomes normal.

Can you help me Mr Boian Mitov?
Your components are perfects! Congratulations!

Best Regards,
Rafael]]>
I am trying use VLDSCapture to capture my webcam video and i linked the output pin to VLHaarDetectObjects (it is configured like the example of facedetect to try detect my face) but when i start the VLHaarDetectObjects the VLImageDisplay is freezing. If i stop the VLHaarDetectObjects, my VLImageDisplay becomes normal.

Can you help me Mr Boian Mitov?
Your components are perfects! Congratulations!

Best Regards,
Rafael]]>