00Zetti's profile - activity

Ok problem solved :D
I could fix it by myself. Like suggested in the comment and what I've already tried - it was the right path. The problem was, that I didn't have the access rights for this folder, now it works fine.

So for others a simple tutorial how to get the files and where to place.

• download and extract pre-trained data
• copy them to /.../opencv_contrib-master/modules/tracking/src/
• make sure you have access rights to this folder (e.g. sudo chmod -R on your opencv Folder)

If you don't need vtk(I am not sure - is it required for basic opencv?) you can build it without, simply tell cmake using a parameter to ignore vtk in build process.

In short: As you should know, the Houghtransformation maps your input image from spatial domain into the Houghspace. You do this with a certain resolution(rho, theta), let's say your accuracy: e.g. you can map your pixels from spatial domain into Houghspace with a stepsize, that defines how often you should change the angle to observe and map the pixels into Houghspace. The smaller the stepsize the more dense your Houghspace will become. Hope that helps you to understand the parameters and the transformation.

One more question: do you know if I need to implement it directly into the library files(like suggested in the documentation) or can I create it as a subclass in my own code? Background: it's part of my bachelor's thesis and should be extensible for future students.
Okay and a second question, would it take much effort to create a scalable tracker since the deadline is coming closer and closer :'D

Okay, I'll make myself familiar with creating a custom tracker and try my best, thanks so far. I've written my own MultiTracker since the default own does not optimize anything and is just simply a kind of a vector, this way I am at least able to use multithreading :D

Yep, that's exactly what I tried :D Maybe there is more to do than just drop these two files? Or do I need to rebuild openCV with the fhe files instead of just placing them afterwards?

Hi,
following the documentation

//In order to use GOTURN tracker, GOTURN architecture goturn.prototxt and goturn.caffemodel are required to exist in root folder.

I have to put the pre trained caffemodel and the goturn.prototxt into the root folder, but which root? I've tried so far:

• openCV src dir (github contrib repo/ + contrib/modules/tracking/+subfolders
• opencv build dir(similar directories here)
• opencv dirs in /usr/local/lib+include+bin
• project src and project build

I am sure there is someone who successfully used it and can help me with this easy question :)

Hi,
I wrote an interface for video labeling and now I want to evaluate the openCV tracking algorithms with a ground truth.
I was looking on the specific params from the trackers, but it seems that only TLD(btw implementation does not perform well) and MEDIANFLOW are able to automatically resize/scale the ROI. Is there a way to obtain an auto-resizing ROI on the other tracking algorithms? Or maybe a simple way to create a custom tracker without changing the code in the lib?