rohitg's profile - activity

Hi! I'm applying for Natural Barcodes project mentioned in the suggested projects list on OpenCV GSoC'14 page. Copied here for reference:

Natural barcodes:
Mentor: Gary Bradski
    OpenCV can read calibration patterns.
        Calibrate a camera,
        rectify it's output.
        Put a picture next to a calibration pattern in a scene.
        Find 2D interest points with features2D.
        Measure them in the image using the calibration pattern.
        Use the pattern of interest points and their measures to turn the picture into a calibration pattern/barcode so that the computer recognizes the picture and knows where it is in relation to the image.
    Stretch goal, embeded and retrieve watermarks in the image.

I am planning to solve it as follows:

1. Compute homography between object image and the scene, using 2d-2d correspondences
2. Compute camera pose (using something like in the answer here)

The obtained camera matrix can then be used to add objects to the scene, with respect to the camera. It can be quite useful for AR applications.

Am I correct in my understanding of the problem and solution. Also, can someone please clarify a bit on the embedding watermarks part? Thanks!

Just to close the question. I sent a PR (https://github.com/Itseez/opencv/pull/2358) and it was accepted :)

Thanks @SpecLad !

Hi! I am trying to build this initializer for RotatedRect as mentioned in a feature request here. Wanted to take feedback before I sent a PR.

Essentially, I take 3 points and derive the center, angle and side lengths using those. I assume the points to be in order (clockwise/anticlockwise), hence center is the mean of first and last points. I find the 2 vectors (corresponding to the 2 edges), and surely one of the vectors must have slope between 1 and -1. That vector (and corresponding edge) becomes our width, and hence the other becomes height. A patch is here.

Please let me know if there are issues with the implementation.

PS: Am really a noob, feel free to pick nits! Thanks!

Hi @StevenPuttemans @Nghia ! well maybe adding these functions would simply overload the API (as per discussion https://github.com/Itseez/opencv/pull/2307 )..however, about the 2nd part of the task, assigning Mat<_tp, n, 1> to Vec<_tp, n>, can it be done someway? or should I implement that operator?

Opened a PR: https://github.com/Itseez/opencv/pull/2307

Thanks for the tip @StevenPuttemans! I actually wanted to do an easy one, but couldn't find an easier one from the Volunteer tasks. If you have one in mind, please let me know!

Hello!

I'm a student and just getting started with contributing to openCV (though have a lot of experience in using openCV). I've already forked, cloned, built and played a bit with the source code. I want to work on the task 3380, and thought it'd be better to run my ideas through other developers before starting to code.

If I understand the question correctly, we essentially want the ::zeros() kind-of initializers for the Vec. In my understanding, Vec is implemented using Mat in modules/core/include/opencv2/core/matx.hpp (line 869). So essentially, I need to add another function there for initialization.

Please let me know if I'm on the right track or not. I'm really a noob in working on open source, so please forgive me for mistakes. Also, I'd really appreciate any guidance on the proper way of collaborating with developers while working on such projects.

Thanks a lot!

Thanks! that works. I also found a sample code for HoG in opencv/samples/cpp/peopledetect.cpp Can a similar setting be done for HAAR classifier?

I wanted to analyze the in-built object detector in opencv (CascadeClassifier, using Haar or Hog features). Is it possible to control the false positive or hit rate of the classifier? I would ideally like to plot an ROC or DET curve of the same.