Disclaimer

This is a personal weblog. The opinions expressed here represent my own and not those of my employer.

19 Responses to Disclaimer

  1. houssam says:

    can i have your paper :
    An empirical evaluation of interest point detectors.

    thankyou

  2. houssam says:

    can i have your paper :
    An empirical evaluation of interest point detectors.
    i did’t find it
    thankyou

  3. houssam says:

    hi
    big thanks for your help

    Best regard

  4. Vitthal says:

    Hi
    I want to Measure size of a box using kinect sensor.

    for example:
    suppose i have CPU box and that box placed in front of kinect sensor…and As a output i want to find all measurement of that box like width ,height and length.

    if you have any solution for same then please help me

    Thanks in advance,
    Vitthal

  5. lu says:

    Hi, I am a PhD student working on Intelligent transportation. Your personal blog helps me a lot on my work. Now I am working on a lane detection programming, and now is looking for some other lane detection methods to compare. Could I have your source code or some other manners to access to your lane detection results using KITTI database, just for result comparison? I’d appreciate it.

    • Hi!
      Thanks for visiting the blog.
      Currently I am working on a lane tracking sample that I will post in the blog for everybody. It is based on the methods described in some of the papers I’ve been publishing.
      I will try to do it soon.
      Kind regards,
      Marcos

  6. lu says:

    Thank you a lot, and pleased to wait for your good news!

  7. lu says:

    Hi, Marcos. I have read your papers on lane marking detection and tracking. In the system evaluation part, you mentioned “ground truth” of a road sequence, e.g. Fig 9 in “Real-time lane tracking using Rao-Blackwellized particle filter”. I am wondering how you get the ground truth lane markings. You marked every marking in every frame by hand, or you use a benchmark provided by some database? Thank you!

  8. plka says:

    Hey, Macros!!
    Can I get any video clip with the top view of vehicles for counting vehicle.

  9. Rafik says:

    Dear Sir,
    I am student at the German University in Cairo, I am preparing my bachelor thesis and I am following your research paper “Video-based Driver Assistance Systems” so first I used the inverse perspective mapping, segmentation, enhancement of the segmentation, blob removal, clustering and now i need to return to the perspective domain but I can not find the code can you help me with it, I am using matlab. Thank you in advance.
    Rafik Noumeir

  10. Malik Al-hallak says:

    Hello Marcos,
    I am a student from germany and currently working an my master’s thesis regarding ADAS. I’ve read your paper: “Road environment modeling using robust perspective analysis and recursive Bayesian segmentation”. I didn’t really understand the lane marking detector part. Could you please explain the algorithm for this?
    Kind regards,
    Malik Al-hallak

    • Hi!
      Maybe you can go directly to sample code snippets I published about it:

      https://marcosnietoblog.wordpress.com/2011/12/27/lane-markings-detection-and-vanishing-point-detection-with-opencv/

      Hope it helps!

      Regards,

      Marcos

      • Malik Al-hallak says:

        Hey,
        yes the code was quite helpfull, thanks. I should have asked more specifically. I have two questions:
        First thing is that because of the large number of black pixels in the lane marker image, I got some problems with singularities in the EM algorithm. Do you have some tips to avoid this, or do you just check sigma_LP/sigma_LO to not turn 0 ?
        My second question is more general. You wrote that you model the unkown group with fix mean and large standard deviation to absorb outliers. How do you determine the values for mean, standard deviation and mixture component for the unkown group for intensity and lane marker feature ?

        Thanks in advance,
        Malik

  11. Malik Al-hallak says:

    Hello Marcos,
    in the meantime I have read the chapter of your Phd and the one of Jon Arróspide Laborda about the bayesian framework. Unfortunatley I was not able to reproduce it. My problem is the “unkown class”. At the moment I have the means and standard deviations for Objects, Pavement and Lanes. I assume that the Unkown class has a mean of 128 (for an 8 bit grayscale image) and a standard deviation of 255.
    Now I would start the em-algorithm. As start coefficients of the mixture model I assume 25% (as I do not have a previous step at the moment). So I start the em-algorithm with 4 Clusters (P,L,O,U) and let it run until convergence is reached. As you wrote, the mean and standard deviation of the unknown class are not updated so that after each m-Step I reset the mean and standard deviation of the unkown class. Is this correct? I have the problem that the unkown class has a very small gaussian mixture coefficient (which is much smaller than 1%) and hence has no really influence on the distribution and won’t absorb the putative outliers. I assume that it’s just a small bit to the correct result. Maybe you could sketch the algorithm a little bit more in detail? Could you please help me with this? I am really interested in using your classifier in my thesis.
    Please feel free to write me a mail directly, if it is okay for you.

  12. Hi,
    If you have problem with the parametrisation of the gaussian component for the outlier rejection, you can probably use the uniforme approximation, i.e. apply your EM on the 3 main classes, and define a ground threshold (the uniforme value) to mark as outlier those samples below a certain likelihood value.
    I don’t remember the details on how we used it, but it is clear that a static gaussian component on the EM causes problems, and must be tackled using the uniform, or resetting the weights associated to the class so that it doesn’t fade away. The important idea is to have a defined threshold, somehow kept constant, to absorb outliers and that is related to your model of the expected likelihood of your samples and the quantity and severity of outliers.
    Hope this helps.
    Regards,

    Marcos

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s