jbhuang0604 / cf2 Goto Github PK

Hello,

This supplement and source are awesome. Thanks a lot.
I am trying to reproduce your paper but blocked.

After running run_tracker, the result is as following and seems wrong .

And command line windows will give the following errors messages occasionally.
Error in tracker_ensemble>predictPosition (line 155)
pos = pos + cell_size * [vert_delta - 1, horiz_delta - 1];

Any suggestions?
Thanks.

I would like to implement your tracking algorithm in ROS and I wonder if you have a c++implementation of.

My matlab is R2014b, matconvnet is beta20.
imagenet-vgg-verydeep-19.mat is downloaded from [(http://www.vlfeat.org/matconvnet/pretrained/)]
I tried to run run_tracker.

So I added net = vl_simplenn_tidy(net); %Amber in initial_net.m
and changed net.normalization.imageSize(1:2) into net.meta.normalization.imageSize(1:2)
as well as net.normalization.averageImage into net.meta.normalization.averageImage in get_features.m

But there still remain some mistakes like this
`Array dimensions must match for binary array op.

Error in get_features (line 17)
img = img - net.meta.normalization.averageImage;

Error in tracker_ensemble>extractFeature (line 201)
feat = get_features(patch, cos_window, indLayers);`

What should I do? Can anyone help me? Thanks!

Hi,

With respect to this video of walking people, could you use this model for tracking multiple people?

If yes, how can you do the same?