ybsong00/vital_release

VITAL: VIsual Tracking via Adversarial Learning

vital_release's Issues

vl_nnconv.m can't be executed

I run the code,but there is some error as following:

try view SCRIPT vl_nnconv as a function to execute:
D:\tracking-matlab\Vital_release-master\matconvnet\matlab\vl_nnconv.m

How can I get the mex file in Windows

Hi, Thank you for sharing good research.
I have a warning about below:
There is no name exist or directory G:\Vital_release-master\Vital_release-master\matconvnet\matlab\mex

Should I make mex file for use?

some error info

when run 'demo_tracking' directly:
err:Attempt to execute SCRIPT vl_nnconv as a function

so I try to re-compile again and run 'compile_matconvnet':
err:No supported compiler or SDK was found. For options

I tried google some methods, but not work. please help~~~~

my enviroment:
ubuntu 16.04
matlab2017A
gcc/g++ 4.9/5.4
cuda 8.0
cudnn 5.1 for cuda 8.0

AUC is about 0.025 lower than yours. Is there anything others to config?

Thank you for your contribution!
I cloned your code and ran directly on OTB-2013, but I got AUC 0.685 and precision 0.903.
Environment: Matlab 2016b & CUDA8.0 CUDNN7 & latest matconvnet (Replace some of your changed files, Because you provide the matconvnet version can not be compiled)
So is there anything others to config? Thank you.

pytorch version

I look forward to your reply about whether the model provided in pytorch version (models/mdnet_vot-otb.pth) can be seen as an official weight, or only models/otbModel.mat in matlab version is certified. Thank you very much.

Loss function

Excuse me，What is λ in the loss function in this paper?

PyTorch version of VITAL tracker

I have implemented the VITAL tracker using PyTorch. It has a higher precision than the MDNet. The url is https://github.com/abnerwang/py-Vital. If there are any problems with my implementation, please feel free to give me feedback.

anybody run successfully?

I run this code and failed,any body can provide the routine to run successfully?

some questions about costsensitiveloss and dropout

It's a great work，but I have a few questions.Can you tell me where the costsensitiveloss is，i searched the whole matlab file，but failed. Also do you mind telling me which code position of feature maps and weight mask dropout?

Implementation in python?

Dear @ybsong00,
Is there any implementation of VITAL in python? or do you have any plan for releasing a python one?

Implementation in python?

关于G网络训练的疑问

在G_pretrain.m文件中代码摘要如下
`prob_k=zeros(9,1);
for k=1:9

row=floor((k-1)/3)+1;
col=mod((k-1),3)+1;

for i=1:nBatches
batch = pos_data(:,:,:,opts.batchSize*(i-1)+1:min(end,opts.batchSize*i));
batch(col,row,:,:)=0;
if(opts.useGpu)
batch = gpuArray(batch);
end
res = vl_simplenn(net_fc, batch, [], [], ...
'disableDropout', true, ...
'conserveMemory', true, ...
'sync', true) ;

f = gather(res(end).x) ;
if ~exist('feat','var')
    feat = zeros(size(f,1),size(f,2),size(f,3),n,'single');
end
feat(:,:,:,opts.batchSize*(i-1)+1:min(end,opts.batchSize*i)) = f;    

end

X=feat;
E = exp(bsxfun(@minus, X, max(X,[],3))) ;
L = sum(E,3) ;
Y = bsxfun(@rdivide, E, L) ;
prob_k(k)=sum(Y(1,1,1,:));

end
[~,idx]=min(prob_k);`

其中**[~,idx]=min(prob_k)**此处选择出来的idx对应的mask，所产生的D网络的loss不是最大的，而是最小的。这样选出的mask与论文中所阐明的选择方法不同。
@22wei22

Is the offline pretraining code available?

Dear Dr.Song,
Thank you for open-sourcing the code. The VITAL paper is an excellent work and reading the paper is an inspiring experience.
I noted that the pretrain folder in this repo does not include the code for the offline pretraining process, as you mentioned in the Model Initialization in Section 4 of your paper. In fact, the folder is identical to that in MDNet and does not involve adversarial training.
I was wondering if you plan to open-source the training process, which I think would be super helpful for the community to understand better how to apply GAN to the tracking problem.
Thanks for your time.
Best regards,
Yiming

寻求帮助

net_G = G_pretrain(net_fc, net_G, neg_data, opts_net);
为什么用负样本初始化G，而不是用正样本初始化G

testing resutls

could you offer some testing results, like OTB100, TrackingNet,VOT2018, UAV...??
thank you!

nnconv_forward_blas<>: : gemm<>: [cublas error] [cublas:cublasSgemm (CUBLAS_STATUS_EXECUTION_FAILED)]

nnconv_forward_blas<>: : gemm<>: [cublas error] [cublas:cublasSgemm
(CUBLAS_STATUS_EXECUTION_FAILED)]
RTX2080ti cuda 8.0 cudnn5.1/cudnn6.1 is the same problem.