he010103 / cfwcr Goto Github PK

未定义函数或变量 'mtimesx'。

I have runned the demo_CFWCR successfully.
But i met errors when running run_experiments below :

In tracker_CFWCR,I use CFWCR_VOT（）.m
And when i use CFWCR_VOT_CPU().m,it could run successfully,but the results using CPU is not as good as GPU.
Do you know how to fix this proble?
Thanks for your reply.

请问 imagenet-vgg-m-2048-cut.mat在哪里下载？

hello.
i had read the code.but i cannot find eq.4 and eq.5

can you tell me where eq.4 and eq.5 are in the code.
thanks in advance.

I can sucessfully run the code demo_CFWCR.m in your test data “sequences/Crossing”.
So I add some new data for further test. I put the image files in the "sequences/frame/img" folder, and a groundtruth_rect.txt in the "sequences/frame".
groundtruth_rect.txt I give only contained one line of coordinate in order tp initialize the program.
And then, the demo_CRWCR can finished without error. However, in the output "results.res" only contained one coordinate, which I gave in groundtruth_rect.txt.
Why the program didn't give some predict result? What was I have missed?

hello ~
could you please upload your results on VOT2016 and OTB100 ? where can i get your results ?

I have runned the demo_CFWCR.m successfully
and i have copied the trax files into the native of VOT file
and then i create the workpace.
But i meet the error when i run run_experiments.m

Tracker execution interrupted: Unable to start the tracker process
TraX support not detected.

Thanks for your reply if you know how to deal with this error.

Hi @he010103 ,
Your method is such good. I want to know that how you train your model? I want to train our own dataset, how I can do? Can you share your training codes?

the install.m has the following line 19
vl_compilenn('enableGpu', true, 'cudaMethod', 'nvcc','cudaRoot', 'your cuda path','verbose', 2);

which has 'your cuda path' for cudaRoot. It is not mentioned in the README that we need to set/change/remove this first before running install

Also, this is a bit unnecessary, since the 2nd step in the README already asks to compile and install matconvnet via the link 'more install details'

How to solve this issue

Error using vl_nnconv
An input is not a numeric array (or GPU support not compiled).

Error in vl_simplenn (line 300)
res(i+1).x = vl_nnconv(res(i).x, l.weights{1}, l.weights{2}, ...

Error in get_cnn_layers (line 23)
cnn_feat = vl_simplenn(fparams.net,gpuArray(im),[],[],'CuDNN',false, 'Mode', 'test');

Error in extract_features (line 48)
feature_map(ind:ind+num_blocks-1) = feat.getFeature(img_samples{img_sample_ind}, feat.fparams, gparams);

Error in tracker (line 275)
xl = extract_features(im, sample_pos, currentScaleFactor, features, global_fparams, feature_extract_info);

Error in testing_CFWCR (line 98)
results = tracker(params);

Error in demo_CFWCR (line 13)
results = testing_CFWCR(seq);

Dear author,
you mention that 'Setting export CUDA_CACHE_MAXSIZE=8000000000" in the ./~bash_profile so that gpuDevice(1) will take fewer time.', and I suppose you run the code in Ubuntu platform.
Could you tell me how to set this in Windows? Thank you!

Best wishes!