Opencv face detection (haar detection)

For my new app I'm in need of a very solid front face detection with haar features written in OpenCV.
Now, I can do this from scratch, but I prefer to purchase this complete module from someone who already did this.
I need detection for: head position, mouth, nose, forehead, eyes regions, eyeballs, hair, lids, brows, lips, mouth corners, teeth, nostrils edges and furrows.
Can somebody help me with that?
~Dennis

I'd suggest you ask in the developer forum, either here or the developer forum you have access to as a member of Apple's Developer Programs (at least, I assume you're a member). You'll be more likely to get suggestions here, if any such module exists. I'm not aware of anything other than the capabilites Apple included in iOS 5 as part of Core Image, but someone else might know of something. There one was independent system - Face.com - but they were purchased by Facebook and have shut down as an independent service.
Regards.

Similar Messages

Face or shape detection?

I work with a magazine that prints mug shots of criminals. I get the mug shots of the people's faces, and I have tons of them to put on my layout. They are all different sizes, and right now I have to go through by hand and resize each one to fit it inside the image frame so that that face fills up the frame and is not cut off. Is there any type of script that could somehow detect the face or shape of the people and automatically size them to fit within the image box? If not, does anyone know an easier way for me to do this? It takes forever because I have hundreds to do.
Thanks!

Curiously, both of the two software packages mentioned so far, SNFaceCrop and ObjectDetect.pm are based on OpenCV, the Open Source Computer Vision library.
Are you using Windows or a Mac? It does matter for some of this stuff. (I guess you must be using Windows because SNFaceCrop is a Windows app; that makes it a bit tougher to assist you, since you need to bring in a lot of extra stuff to run perl or python or other scripting languages that seem to have easily accessible OpenCV bindings...)
I think what you are asking for is fairly outside the realm of what can be reasonably expected to be provided on this forum. So someone may pop up and be very generous and invest the time to give you a solution, but I'm afraid you are better off breaking the problem into two parts and asking elsewhere.
The part you should solve elsewhere: feed a list of files to a program that invokes the OpenCV library and determines the face position and produces a list of files and their face positions.
Then, given that, we can show you how to take that list into InDesign and do {whatever you want}. Crop the photos to the face positions, leaving some margin, array the images in some intelligent fashion based on the face positions, etc., etc.
That said, I'm not sure what the best place to ask is. I would (again), probably suggest starting in the PhotoShop Scripting forum, though in truth that is probably a bad suggestion: it is less active than this one and you don't really want to solve the problem in PhotoShop. Still, you want to reach out to some people who regularly think about image processing...
You might also spend a while googling or OpenCV-based projects (seems like there are a lot) and trying to find one that does what you want. It probably exists, it sounds like a straightforward application of that software.
GOOD LUCK! Please do report back here, whatever you find!

Faces not auto-detecting

Faces frequently fails to auto-detect faces in photos. I have several pictures where the initial scan produces no faces to be named, even when the faces in the pictures are front-on and unobscured.
On the other hand, I frequently get 'faces' auto-detected on things like peoples arms, legs, speaker grills, etc.
Is this an indication of a problem with Faces, or is it par for the course?
Thanks!

Thanks, that was what I was afraid of. In general, Faces is excellent. As you said though, it is far from perfect.
Dan

Face Recognition and Detection

I did quite a bit of research on the topic and found several ways of face recognition:
-eigenface which is basically an average face from several photos and then faces are compared to it
problem is that the pictures must all be just of the face, be black and white (color is harder) and have same lighting
this is impossible in the real world
The way I was thinking is to find the coordinates of the eyes and nose, and based on the distance from eye to eye, and R eye to nose, L eye to nose base my recognition (those values should be unique for each face)
I found this thing called opencv which was made by intel, and apparently using haarcascades a lot of people have succesfully been finding the coordinates of eyes and nose in images (images that are of people, not just an ideal photo of a face)
The problem is that opencv is written in C, so to use it (the rest of my app is in java), I would have to use JNI or something similar.
Can anyone help me out with that? Or maybe there is a way to do it purely in java?
Thanks a lot for your help

If you want to integrate a C/C++ application from JNI into Java this may be helpful :
http://www.javaworld.com/jw-03-2001/jw-0316-itw-jni.html
html format :
http://java.sun.com/developer/onlineTraining/Programming/JDCBook/jni.html
pdf format:
http://java.sun.com/developer/Books/j2ee/advancedprogramming/jni.pdf
Edited by: aurelian_cl on Mar 6, 2008 1:09 PM

I have a mac book air 11". When I boot up connected to the huge Thunderbolt screen it detects the detects the fact and gives me great quality. now I am travelling away from the scree it boots up, nice picture, then switches to rubbish resolution.

I have a Mac Book air 11". Normally at home I connect this to a large Thuderbolt Screen. When booting up it some times gets confused and offers me a low resolution result, but some plugging and unplugging of the screen fixes this. I assume it sometimes "sees" the big screen and sets the resolition accordingly. Sometimes does not. But now... I am away from big screen and cannot get the Macbook air on its one to deliver anything other than really poor resolution. I ahve done displays and chosen all three - but no better. And I have said Choose the Best ONe... again not luck. Any ideas. I read in back posts this was a problem a few years ago.

Welcome to Apple Discussions
I believe you have a failing GPU (aka Video Card). Run AHT and see what it tells you. Do not be afraid to run it up to three times in succession if at first it does not detect an error (sometimes errors that are caused by overheating will not show up until the computer is stressed enough - repeating AHT will do that).
As an aside - the MBA was not designed to take the place of a desktop. If you are powering your MBA by leaving it permanently attached to the thunderbolt port on the monitor you are shortening your battery life.

Internal display not detected. "Detect displays" option now gone...

Hello!
My MacBook Pro will not detect any displays. After going to sleep a couple of nights ago, and bringing the computer out of sleep, the display would not turn on; even though I could hear the computer wake up and the fans turn on.
I hard-reset the computer and the display still would not turn on after the power down. Was able to connect to my computer through my wife's MacBook through "Share Screen", and everything is working except the display. I immediately backed up my data realizing the system was working fine (not a hard drive issue, hooray!) other than the display. I tried connecting my computer to an external display and did a restart with the lid closed and the computer would not detect the external either.
Through share screen, I then went into the Display Properties and to my surprise noticed that the "Detect Displays" button has disappeared! Furthermore, when I go into
"About This Mac > More Info... > Graphics / Displays"
Nothing is listed for the screen! On my wife's MacBook it lists the LCD's resolution and other information, but not on mine. Also, under "Display Connector", hers says "Status: No Display Connected"; where as mine says nothing at all, its blank underneath the words "Display Connector".
I reset the P-RAM and that did not help either. If I could I would do a fresh install of OSX, but I can not see the screen so I can do that either.
Is there any way I can reset the computer by any other means?
I also don't have an issue with taking this computer apart, as apple care is expired, and I've done my fair share of take aparts before.
Once again, the computer is starting up and working, but the screen is not being detected or turning on. Everything else works. I called apple and they wanted me to pay $50 for an over the phone troubleshoot, or to take it in to the apple store (an hour away from here) and possibly pay $300 for it to be fixed. AHHHH!
Thanks.

OK, I am confused...can you see an image on an external display or not?
The torch/flashlight trick will tell you if the backlight is out and you are still getting an image and it is a display problem. Although dead backlights are less of a problem with the LED backlit displays.
In your last post you say "If I could at least detect an external...", this leads me to believe that you are not able to detect an external display. With this assumption, it looks like the video card, which if memory serves me correctly is on the MLB and would require replacing the MLB.
If you are able to get an image on an external display, then we are back to the display as the problem.

Anomaly Detection not detecting host machines (learned OS)

I have an ASA5540X firewall with the internal (software based) IPS module. The module has the up-to-date signatures and seems to be running correctly. However, after enabling anomaly detection (ad0), and specifying the internal zones, I don't see any "Learned OS" in IME
My settings are pretty basic for the sensor
access-list ips_traffic extended permit ip any any
access-list ips_traffic extended permit udp any any
class-map ips_class
match access-list ips_traffic
policy-map global_policy
class ips_class
ips inline fail-open
not sure why it isn't learning the OSs

Learned OS maps—OS maps observed by the sensor through the fingerprinting of TCP packets with the SYN control bit set. Learned OS maps are local to the virtual sensor that sees the traffic.
can you verify the OS finger printing from
sensor# show os-identification learned
Enable passive-traffic-analysis {enabled | disabled}

Detecting folder & detecting if entire contents are there

I've seen some old discussions about detecting if a folder exists and a
reference to an enhancement request for detecting folders in bundles.
Did that every get completed and does it include detecting if an entire
folder matches another folder before continuing?
Thanks.
Delon

You can use the "File Exists" option for Folders as well as Files.
There is not any mechanism to compare an entire folder structure in
System Requirements.
On 7/3/2014 11:28 AM, Delon Weuve wrote:
> I've seen some old discussions about detecting if a folder exists and a
> reference to an enhancement request for detecting folders in bundles.
>
> Did that every get completed and does it include detecting if an entire
> folder matches another folder before continuing?
>
> Thanks.
>
> Delon
>
Craig Wilson - MCNE, MCSE, CCNA
Novell Technical Support Engineer
Novell does not officially monitor these forums.
Suggestions/Opinions/Statements made by me are solely my own.
These thoughts may not be shared by either Novell or any rational human.

Can there be a N79 Camera Update with Face Detecti...

Hello there, I'm yet to have this phone delivered but I'm looking forward to it. I was thrille dto hear that Nokia do updates which can also enhance this phone's cameras which is really cool. My question is if Nokia could make an update that gives the phone Face Detection, Smile Detection, and all those other modern camera features? It seems unlikely to me but I was just wondering if they could do this, and if they, can they do it soon!
Thanks.

well the nokia n86 has this feature who knows it may be in the line down the road you can use the contact us link and suggest it to nokia there
If i have helped at all a click on the white star below would be nice thanks.
Now using the Lumia 1520

How to not use Face naming to help detect further faces?

HI,
My question may sound funny.
My understanding is both iPhoto and Aperture use detected face in helping detect other faces. That's what the help suggest, in identifying more face it facilitate the detection.
In many photo you would have someone visible and would want to identify her/him. by doing so I would use anything but the face, the back of the head or other part of the body that is visible.
If Aperture/iPhoto would then use that added image to improve face detection with nothing useful, it would make it even less efficient!
Are their other way to identify someone and have it in the Faces without affecting the faces definition the the software seem to be building? tag not to use the face for further detection for example.
Vincent

See my answer to your duplicate question.
Re: How to not use Face naming to help detect further faces?
Naming body parts will not hamper the face recognition algorithm. iPhoto will ignore manually added faces, see this link:
iPhoto '09 & iPhoto '11: Improving face recognition results

Detect Missing Faces

I recently learned that through an update a new feature was added to Faces. When you right-click on a picture in the iPhoto library you can now select Detect Missing Faces. This feature apparently relaxes the face detection algorithm so that faces that were not found in the initial scan (or during the upload) may now be detected. Unfortunately, this feature is only available when one is not in Faces and you must be looking at the thumbnails.
In any case, it does not seem to work for me. I have several pictures with clear head-on shots where the faces were not detected (and I don't mean recognized); when I exit Faces and right-click the image to select Detect Missing Faces, nothing happens.
Is it working for anyone? Is there a trick?
Thanks.

I have the same frustrating problem. Everytime I add new pictures, faces is not recognizing it. I would have to individually tag them, even though I already have hundreds of pictures with the same face in my corkboard. Help
jr

Aperture 3.0 and 3.0.1 Face Detection

In general I have been very pleased with Aperture 3.0. I have updated several large libraries to 3.0 without incident. However Aperture face recognition and detection is very buggy, even with the 3.0.1 update. Aperture 3.0.1 still has major problems with face recognition, including false detection of faces outside of the picture area, readily visible when zooming out. It needs a better way to delete false face recognition than simply clicking on the checkbox at the top left hand corner of the face area, which may well be outside of the picture area. Many photos still indicate they have unnamed faces, even with none visible even when zoomed out.

Having just upgraded to 3.0.1 and having successfully begun categorizing my photos with 'Faces', I tried to use the 'Name' button on some photos and was very surprised to see that the face positioning boxes were in the wrong places on the photo. When I added a new face box, properly positioned, going into the 'Faces' view showed...leaves...or clothes, or cars, or just about anything OTHER than the face I had boxed.
That's when I found this discussion area, being sure that I couldn't be the only one with this issue. And I wasn't.
THEN...I noticed that lots of my photos were 'offline' from their referenced files...in fact, almost every one of them was offline (all my photos are referenced). I went looking and figured out that when I had replaced my internal HDD +and had given it a different name+, Aperture couldn't find them anymore. A few minutes of 'Reconnect All' and now the 'Name' feature is working much better.
I'm sure this won't address the issues you more advanced users have reported, but...should anyone else be working with photos with offline referenced files, you might try reconnecting them.

Mouse detecting object component

I'm writting an applet in which there're 3 objects(same class) on the screen, when I clicked and dragged the object, it'll move along with the mouse. I'm having trouble to let the mouse recognise which object it is clicked and dragged. Can anyone please help me? I'm new to Java and I need to slove this problem as soon as I can. Thanks!

You didn't say what kind of objects you are trying to move. Looking at your earlier threads it appears you were working with graphics (as opposed to components like Panels or Labels). Here's an example of detecting and moving graphic shapes in the AWT. Swing is a little easier to do this in because of the double buffering. I made it so you can run it in either j2se 1.4 or the newer 1.5 (which requires generics).
import java.awt.*;
import java.awt.event.*;
import java.awt.geom.*;
import java.util.*;
import java.util.List;
import javax.swing.event.MouseInputAdapter;
public class Detection
    public Detection()
        DetectionPanel detectionPanel = new DetectionPanel();
        Detector detector = new Detector(detectionPanel);
        detectionPanel.addMouseListener(detector);
        detectionPanel.addMouseMotionListener(detector);
        Frame f = new Frame();
        f.addWindowListener(new WindowAdapter()
            public void windowClosing(WindowEvent e)
                System.exit(0);
        f.add(detectionPanel);
        f.setSize(400,400);
        f.setLocation(200,200);
        f.setVisible(true);
    public static void main(String[] args)
        new Detection();
class DetectionPanel extends Panel
    List <RectangularShape> shapes;                                     // j2se 1.5
    // List shapes;                                                     // j2se 1.4
    boolean firstTime;
    public DetectionPanel()
        shapes = new ArrayList<RectangularShape>();                     // j2se 1.5
        // shapes = new ArrayList();                                    // j2se 1.4
        firstTime = true;
    public void paint(Graphics g)
        super.paint(g);
        Graphics2D g2 = (Graphics2D)g;
        g2.setRenderingHint(RenderingHints.KEY_ANTIALIASING,
                            RenderingHints.VALUE_ANTIALIAS_ON);
        if(firstTime)
            initShapes();
        for(int j = 0; j < shapes.size(); j++)
            g2.draw(shapes.get(j));                                     // j2se 1.5
            // g2.draw((RectangularShape)shapes.get(j));                // j2se 1.4
    private void initShapes()
        int w = getWidth();
        int h = getHeight();
        int dia = Math.min(w,h)/4;
        shapes.add(new Rectangle(2*w/3, 2*h/3, dia, dia));
        shapes.add(new Ellipse2D.Double(w/4 - dia/2, h/3 - dia/2, dia, dia));
        firstTime = false;
    public void moveShape(int index, int x, int y)
        RectangularShape rs = shapes.get(index);                        // j2se 1.5
        // RectangularShape rs = (RectangularShape)shapes.get(index);   // j2se 1.4
        Rectangle clip = new Rectangle();
        clip.x = (int)(x < rs.getX() ? x : rs.getX());
        clip.y = (int)(y < rs.getY() ? y : rs.getY());
        clip.width = (int)(clip.x + rs.getWidth());
        clip.height = (int)(clip.y + rs.getHeight());
        rs.setFrame(x, y, rs.getWidth(), rs.getHeight());
        repaint(clip.x, clip.y, clip.width, clip.height);
    public List<RectangularShape> getShapes()                           // j2se 1.5
    // public List getShapes()                                          // j2se 1.4
        return shapes;
class Detector extends MouseInputAdapter
    DetectionPanel detectionPanel;
    int selectedIndex;
    Point offset;
    boolean dragging;
    public Detector(DetectionPanel dp)
        detectionPanel = dp;
        offset = new Point();
        dragging = false;
    public void mousePressed(MouseEvent e)
        Point p = e.getPoint();
        List<RectangularShape> list = detectionPanel.getShapes();       // j2se 1.5
        // List list = detectionPanel.getShapes();                      // j2se 1.4
        for(int j = 0; j < list.size(); j++)
            RectangularShape rs = list.get(j);                          // j2se 1.5
            // RectangularShape rs = (RectangularShape)list.get(j);     // j2se 1.4
            if(rs.contains(p))
                offset.x = (int)(p.x - rs.getX());
                offset.y = (int)(p.y - rs.getY());
                dragging = true;
                selectedIndex = j;
                break;
    public void mouseReleased(MouseEvent e)
        dragging = false;
    public void mouseDragged(MouseEvent e)
        if(dragging)
            int x = e.getX() - offset.x;
            int y = e.getY() - offset.y;
            detectionPanel.moveShape(selectedIndex, x, y);
}

IE10 setup detection method

Installing IE10 in Task sequence. Not using IEAK, so command line is
IE10-Windows6.1-x86-en-us.exe" /quiet /update-no /norestart
Installs fine and reboot suppressed. Problem is the detection method detects as failed because prior to reboot nothing appears to change. Neither iexplore.exe or the registry entries have been updated to V10. Therefore SCCM registers as a failed install
After reboot all fine, but the reports say that step of the TS failed
Is there a way round this ?
Ian Burnell, London (UK)

Ian,
I would recommend that you use packages for OSD rather than applications. You don't need detection with OSD, it's completely irrelevant and if used within the TS those rules need to be valid. Create an application for non OSD endpoints that are already out
there that need the update.
Packages give you the option also to run direct from the DP - which might be handy for larger applications that fill up the cache locally.
Ultimately keep it simple. Why over complicate your OSD with unnecessary 'frigging' around?
Just because the model exists doesn't mean you are reducing to using that method only.
Cheers
Paul | sccmentor.wordpress.com

Complex collision detection: sides of a cube

I was wondering if it is possible with AS3 to use complex collision detection, for detecting which side of a cube is hit by an object.
I know i can also split up the cubes sides and write different hittests for them, but that seems like writing unnecessary code...

Either use Photoshop’s 3D features or use three instances of a Smart Object that you manually transform accordingly.

Opencv face detection (haar detection)

Similar Messages

Maybe you are looking for