For discussion on techniques for aqcuiring and analysing images and other high dimensional data in order to produce information.

• Gang Zhang

  Which method is better for eye detection, AdaBoost, or SVM?

  I have used AdaBoost to detect a face. When I used AdaBoost to detect an eye, I found that detection performance is low. It is written in some papers that SVM was used to detect an eye after AdaBoostI have used AdaBoost to detect a face. When I used AdaBoost to detect an eye, I found that detection performance is low. It is written in some papers that SVM was used to detect an eye after AdaBoost was used for face detection. Whether it mean that SVM is better than AdaBoost for eye detection. Besides, whether whole detection of eye and brown is better than only eye detection.

  Recent replies ⋅ Show All (2)

  • 
    Nikolay Sergievskiy replied

    Hi If you whant to choose from HoG SVM (from pedastrian detector) or Haar AdaBoost (ViolaJones algorithm); I recomend you use AdaBoost and create some penalty function (that puts eye position intoHi If you whant to choose from HoG SVM (from pedastrian detector) or Haar AdaBoost (ViolaJones algorithm); I recomend you use AdaBoost and create some penalty function (that puts eye position into face boundary box).

    20 days ago
• Nikolay Sergievskiy

  Learning and defining subclasses in image annotation

  In many cases I want to discriminate my data to subclasses; 1. If I want to detect 'chair' I want to know further if it is: sofa, office chair, kitchen chair.... (LabelMe has just the tag chair). 2.In many cases I want to discriminate my data to subclasses; 1. If I want to detect 'chair' I want to know further if it is: sofa, office chair, kitchen chair.... (LabelMe has just the tag chair). 2. If I want to detect 'car' I want to categorize cars by shape and sight automatically (In modern LatentSVM they do it by heuristics). 3. If I want to detect 'car' by color I want to have a guaranteed detection, that means I need red, green, etc and I dont care about gray cars. Some problems may be solved by simple categorization by Euclidian measures, but I am interested in Non Linear Clusterization (or latent subclasses variables) that minimize main classification error. Are there any options for more accurate classification?

  Recent replies ⋅ Show All (5)

  • 
    April Khademi replied

    Perhaps you could start with supervised classification - i.e. using training data and the leave one out method (LOOM) to train and classify. Then you could see what clusters your features form inPerhaps you could start with supervised classification - i.e. using training data and the leave one out method (LOOM) to train and classify. Then you could see what clusters your features form in the feature space and whether the features are robust enough to discriminate between the subtypes in the first place.

    Apr 23, 2012
• Julian Mahecha

  Object Recognition

  I'm starting my research in object recognition, I'm doing an initial state of the art of computer vision and object recognition. Someone could suggest me some trends or techniques for furtherI'm starting my research in object recognition, I'm doing an initial state of the art of computer vision and object recognition. Someone could suggest me some trends or techniques for further research?

  Recent replies ⋅ Show All (5)

  • 
    Kadu Vido replied

    If you don't have experience in Computer Vision, I suggest you watch the video classes on Coursera: https://www.coursera.org/course/vision Although there's no way to know for sure how good theseIf you don't have experience in Computer Vision, I suggest you watch the video classes on Coursera: https://www.coursera.org/course/vision Although there's no way to know for sure how good these will be, Professor Malik is known in the field.

    Apr 17, 2012
• Fatima Akhmedova

  What is the best C# wrapper for OpenCV library?

  I learn OpenCV, but as I am more experienced in C# than C++, I am thinking of using wrapper. Up to now, I've tried two of them: OpenCVSharp and Emgu СV and the last one seems to be better. What canI learn OpenCV, but as I am more experienced in C# than C++, I am thinking of using wrapper. Up to now, I've tried two of them: OpenCVSharp and Emgu СV and the last one seems to be better. What can you suggest?

  Recent replies ⋅ Show All (5)

  • 

    Comment protected

    Join ResearchGate now to read this comment.
• Johnny Cunningham

  What's everones thoughts on multiple compression layered programming, also using a virtual disk to run parellel

  Theory
• Iñaki Cormenzana Méndez

  Why two frame grabbers digitizes the same color on different values ​​in memory? My App does not process one of them.

  Hello, at the "Universidad Nacional de Tucumán, Argentina - Depto. Luminotecnia, Luz y Visión", we are developing a Portable Eye tracker. Recently we bought a frame grabber sensoray model 2255 toHello, at the "Universidad Nacional de Tucumán, Argentina - Depto. Luminotecnia, Luz y Visión", we are developing a Portable Eye tracker. Recently we bought a frame grabber sensoray model 2255 to migrate our current frame grabber from the company http://www.theimagingsource.com the model is DFG/LC2 (PCI). The App is capturing the image of the eye with a monochrome camera in NTSC mode and COLOR_Y8 - monochrome image, 1 byte per pixel. The issue is that for some reason,with the frame grabber Sensoray, my App does not process the image, digitized with success but when the "Class Processor" of my software tries to find the pupil, that is, interpret the image, does not succeed (a process that performs well with the old frame grabber. Apparently (from what I understand), the problem is how to represent the image in memory. I did a test with the two frame grabber, putting a totally black image in thecamera, and the result in the old frame grabber was an array of all values ​​"00" in hexadecimal, but with the frame grabber sensoray other values ​​are obtained near to "20" in hexadecimal clear. Even setting the brightness and contrast to a minimum, the smallest value represented is "01" in hex for sensoray frame grabber. Why the frame grabber sensoray does not represent the black with "00"? I'm missing something? I would greatly appreciate if you can help me. Best Regards,

  Recent replies ⋅ Show All (5)

  • 
    Iñaki Cormenzana Méndez replied

    Keivan, I researched a bit more .. For some reason that still do not understand the engineers of Sensoray decided that the 2255 (Frame Grabber) output range is 1 to 254 (0xfe) (at extreme brightnessKeivan, I researched a bit more .. For some reason that still do not understand the engineers of Sensoray decided that the 2255 (Frame Grabber) output range is 1 to 254 (0xfe) (at extreme brightness and contrast settings). The issue is that the App algorithm that segments the pupil (among others things), searches for a string of intensity values ​​"0"!, So the algorithm never find anything because what comes out is at least 1: S. Besides this, the frame grabber has a kind of Dark Noise. For example, I did an experiment covering the camera lens with a black image on the scanned image and calculated the average of these values ​​in black, the result of average is between 25 and 28 Hex! and the maximum can be 31: S. The issue now is, how remove that noise? Should refurbish the image before sending it to the processor class?. Or modify the algorithms making it tolerant to noise?, Mmm I do not like this last idea. I am researching a bit .. Thanks for the replies.

    Dec 12, 2011
• Yuanhao Gong

  OpenCL for Computer Vision(CLCV)

  I am trying to develop a library based on OpenCL for computer vision using c++. The basic assumption here is that images are sparse and pixels can be point wise handled. The second level is toI am trying to develop a library based on OpenCL for computer vision using c++. The basic assumption here is that images are sparse and pixels can be point wise handled. The second level is to develop variational methods on it. Contact me if you are interested: ygong@inf.ethz.ch
• Pinki Sharma

  Gender detection

  Can anybody tell me about gender classification by face?

  Recent replies ⋅ Show All (14)

  • 
    Pinki Sharma replied

    sir i sent the chat requst on gmail plz accept

    Jan 23, 2011
• Rahul Gaurav

  Codes needed__ AnNy suggestions

  Hi folks.......... I am looking for programming codes for extracting videos from any external device (say webcam, CCTV etc)... Could anyone help me out ... Thanks in advance :)

  Recent replies ⋅ Show All (5)

  • 
    Prabhu Marimuthu replied

    hi need to run OpenCV on ARM/Dsp processor can u please give some suggestions...

    Jan 13, 2011
• Faizan Asif

  Blob Detection

  Can any one tell me about writing a simple blob detection algorithm in Matlab for circular object detection?

  Recent replies ⋅ Show All (1)

  • 

    Comment protected

    Join ResearchGate now to read this comment.
• Zahra Sadeghi

  object recognition

  Can anyone tell me what are the hot topics in machine vision about object recognition using machine learning methods?

  Recent replies ⋅ Show All (2)

  • 
    Zahra Sadeghi replied

    Thanks! Do you know what is the very recent advancement and the hot challenging problems related to it? Also, I need some good survey articles about the new machine learning methods used in objectThanks! Do you know what is the very recent advancement and the hot challenging problems related to it? Also, I need some good survey articles about the new machine learning methods used in object recognition.

    Nov 2, 2010
• Apoorvagiri L D

  Image processing

  Face emotion detection using computer vision regarding

  Recent replies ⋅ Show All (6)

  • 

    Comment protected

    Join ResearchGate now to read this comment.
• Burak Çizmeci

  Computer Vision Lecture Notes

  These are the links for the lectures notes on Computer vision, I have also added these links to ''Files'' sectionThese are the links for the lectures notes on Computer vision, I have also added these links to ''Files'' section above. http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT1/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT2/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT3/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT4/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT5/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT6/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT7/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT8/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT9/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT10/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT11/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT12/node1.html http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT13/node1.html Good Luck

  Recent replies ⋅ Show All (2)

  • 
    Burak Çizmeci replied

    you welcome my colleague, I will share additional sources soon. by the way, lets merge the photos ;) but where is the man in the middle

    Oct 1, 2010