vl_nnreshape_wzhshi.m

function y = vl_nnreshape_wzhshi( x, dims, dzdy )
%VL_NNRESHAPE_WZHSHI Summary of this function goes here
%   Detailed explanation goes here

% VL_NNRESHAPE Feature reshaping
%   Y = VL_NNRESHAPE(X, DIMS) reshapes the input data X to have
%   the dimensions specified by DIMS. X is a SINGLE array of 
%   dimension H x W x D x N where (H,W) are the height and width of 
%   the map stack, D is the image depth (number of feature channels) 
%   and N the number of of images in the stack. DIMS is a 1 x 3 array
%   of integers describing the dimensions that Y will take (batch 
%   size is preserved). In addition to positive integers, the 
%   following can also be specified in the style of caffe:
%
%   Interpretation of DIMS elements:
%   -1 := work it out from other dims
%    0 := copy dimension from X
%
%   NOTE: At most one dimension can be worked out from the others.
%
%   DZDX = VL_NNRESHAPE(X, DIMS, DZDY) computes the derivatives of the
%   block projected onto DZDY. DZDX and DZDY have the same dimensions
%   as X and Y respectively.

assert(sum(dims == -1) <= 1, 'at most one dim can be computed from the others') ;
assert(length(dims) == 3, 'dims should have three elements') ;

sz = size(x) ;

copyDims = find(dims == 0) ;
if copyDims
    dims(copyDims) = sz(copyDims) ;
end

targetDim = find(dims == -1) ;
if targetDim
    idx = [1 2 3] ;
    idx(targetDim) = [] ;
    dims(targetDim) = prod(sz(1:3)) / prod(dims(idx)) ;
end

dims = horzcat(dims, size(x,4)) ;

if nargin <= 2 || isempty(dzdy)
    y = reshape(x, dims) ;
else
    y = reshape(dzdy, size(x)) ;
end