"unwrap" function

[mjabri]

I looked at the python bindings and could not find the "unwrap" function which is supposed to implement MATLAB im2col like op. Did i miss it? If not are there plans to have it?
What i am trying to do is a convolution of an image with multiple kernels using a matrix multiplication approach as for example in Caffe. For this, the image needs to be 'lowered' (ie using an im2col like op) so each extracted patch become (raveled) row in a matrix, and the kernels are each a column in the second matrix. After the matrix multiplication, the resulting matrix has to be converted back using a dual 'col2im' op where each column is converted back to an image (convolution output corresponding to same kernel column).

[pavanky]

This will done once it is implemented in C++

[mjabri]

You mean the "wrap" (col2im) part? Because the im2col part ("wrap") is what is implemented in c++ already.

[mjabri]

Ok, i see the confusion. There are two issues, and i really should have raised them separately. So what I am raising here is whether unwrap will be put in the python bindings soon.

[pavanky]

@mjabri The "unwrap" function has been added to ArrayFire in the devel branch after 3.0 release. The new functions added during the 3.1 development cycle will be added to the python bindings after 3.1 is released.

[pavanky]

This is now present in the master branch.

[mjabri]

Thanks!!