//--------------------------------------------------------------
ofVec3f tensor_to_pixel_dims(const tensorflow::Tensor &t, string chmap) {
int rank = t.shape().dims();
vector<int> tensor_dims(rank);
for(int i=0; i<rank; i++) tensor_dims[i] = t.dim_size(i); // useful for debugging
// add z to end of string to top it up to length 3, this'll make it default to 1
while(chmap.length()<3) chmap += "z";
// which tensor dimension to use for which image xyz component
// initially read from chmap parameter
ofVec3f dim_indices(chmap[0]-'0', chmap[1]-'0', chmap[2]-'0');
// if tensor rank is less than the chmap, adjust dim_indices accordingly (
if(rank < chmap.length()) {
if(rank == 1) {
// if(dim_indices)
dim_indices.set(0, 99, 99); // set these large so they default to 1
} else if(rank == 2) {
if(dim_indices[1] > dim_indices[0]) dim_indices.set(0, 1, 99);
else dim_indices.set(1, 0, 99);
}
}
ofVec3f image_dims (
(rank > dim_indices[0] ? t.dim_size( dim_indices[0]) : 1),
(rank > dim_indices[1] ? t.dim_size( dim_indices[1]) : 1),
(rank > dim_indices[2] ? t.dim_size( dim_indices[2]) : 1)
);
return image_dims;
}
InputParameters validParams<FunctionGradAux>()
{
InputParameters params = validParams<AuxKernel>();
MooseEnum dim_indices("x=0 y=1 z=2", "x");
params.addRequiredParam<FunctionName>("function", "Function used to compute gradient");
params.addParam<MooseEnum>("dimension_index", dim_indices, "The dimension index x|y|z");
return params;
}
