vojirt / kcf Goto Github PK

I cmake kcf_vot and kcf_trax. but when I ran kcf_trax, it kept waiting at this line.
Is there any guide or suggestions?

Hey Tomas,

I am translating a MATLAB algorithm to C++ using OpenCV and I was hoping to use your FHOG C++ wrapper. However, I am wishing to distribute my work under the BSD license. Is it possible to get a version of your wrapper under the BSD?

Thank you for the consideration.

I have done as your advice ,but fail to run, maybe the line:VOT vot; // Initialize the communcation is the bug????

please somebody tell me where can i get the images.txt ? Thanks a lot!

Hi:

I have the problem with time spent when I trying to use this KCF tracker in Opencv2
Although high accuracy of this tracker, some time the KCF part need more than 100 ms, even 200 ms

Is there any solution or adjustable parameter to be tuned to speed up this tracker

Thanks!

I have complied under Linux, ./kcf_vot ,but I do not know how to right the format of input file , could you give me some example？Thanks

great project !! i just want to know if there is any way to speed up the tracking. running on arm cpu, it cost almost 2 sec per frame.

A bug in gradientMex.cpp:

R = (float*) wrCalloc(wb*hb*nOrients,sizeof(float));
wrFree(R); 

N,R1, R2 also have the same problem, someone can modify like this:

R = (float*)alMalloc(wb*hb*nOrients*sizeof(float),16);
memset(R, 0, wb*hb*nOrients*sizeof(float));
alFree(R);

Will work correctly.

What is trax? Where can I download it and have it installed?

Thanks...

I fail to run:
./kcf_vot region.txt images.txt
the region.txt have 4 numbers:246,226,94,114,
the images.txt is the absolute path of images:/home/daixiongming/下载/kcf-master/build_dai/blurFace/img.

the blurFace is downloaded on visual tracker benchmark!

but I also fail to run!!!!could you give me some advice ????thanks.

you will also see the build_dai directory include two files(images.txt and region.txt), and ./kcf_vot ,it also fails to read these files, could you tell me how to solve this problem? Many thanks! many friend in our group also fails to solve it !!!!

thank you for your code! it is nearly 1fps when i use the code. how to speed up to make real time?

I run the code 2 times(one using linear kernel, another using gaussain kernel) in OTB50, every setting is same except the kernel. The results are ,the tracking speed of using linear kernel is 236fps, and the tracking speed of using gaussain kernel is 234fps.
In theory, the speed of using linear kernel is faster than using gaussain kernel.

Chinese government refuse to review foreigner website, your email address have some fault words, could you give me your address on this page?? Many thanks!

Had anyone used fftw to replace dft experience?
thanks for help.

Dear Tomas,
Many thanks for your sharing about the code, below is my running about "the kcf"
could you give me so advice ？ I have spend 2days , and I also fail to run!

My error message is like this:

./kcf_vot
init: img size 640 480
init: win size. 336 172
init: min max scales factors: 0.11781 2.20804
kcf_vot: /home/ubuntu/Tracker/vojirt_kcf/src/complexmat.hpp:149: ComplexMat_ ComplexMat_::mat_mat_operator(void (*)(std::complex<_Tp>&, const std::complex<Tp>&), const ComplexMat&) const [with T = float]: Assertion `mat_rhs.n_channels == n_channels && mat_rhs.cols == cols && mat_rhs.rows == rows' failed.
Aborted (core dumped)

I used the tiger in vot-2016.

~/Tracker/KCF/vojirt_kcf$ head region.txt
8,29,134,69

~/Tracker/KCF/vojirt_kcf$ head images.txt
~/Tracker/vot2016/tiger/00000001.jpg
~/Tracker/vot2016/tiger/00000002.jpg
~/Tracker/vot2016/tiger/00000003.jpg

Do you have clue?

BTW, does the code require Matlab library?
Thanks a lot for your help.

Hi，I wonder the principle of the function " cv::Point2f KCF_Tracker::sub_pixel_peak(cv::Point & max_loc, cv::Mat & response) " in track.cpp. How does it work? It seems that there is no this step in the original paper. Many thanks!

gradient.cpp
void fhog( float *M, float *O, float *H, int h, int w, int binSize,int nOrients, int softBin, float clip )

hello, how to use linear kernel ? can you give some advises?

looking forward your reply.

Hi,
Thank you for your code. In the "track" function in "kcf.cpp", shouldn't you first find "p_current_scale" and then "p_pose"?


As shown above, I mean the code in the figure of "find p_current_scale" should be before the code in the figure of "find p_pose".

If I wanna run this project with XCode by c++，how chould I do

there are some question about this:

1. can I run without trax?
2. whether vot must need matlab to run?
3. the file "gen_cpp_cn.m" need to change to c++ file or something else?

thanks for help.