OpenCV YAML without knowing node names

Question

Can I parse a YAML file in OpenCV (using FileStorage library) without knowing the names of the nodes?

For example, the next YAML file:

%YAML:1.0
descriptors1: !!opencv-matrix
   rows: 82
   cols: 64
   dt: f
   data: [ ... ]
descriptors2: !!opencv-matrix
   rows: 161
   cols: 64
   dt: f
   data: [ ... ]

and parse, node by node (OpenCV Matrix by OpenCV Matrix), without doing:

FileStorage fs;
...

Mat firstMatrix;
Mat secondMatrix;
fs["descriptors1"] >> firstMatrix;    
fs["descriptors2"] >> secondMatrix;

Answer 1

To obtain all keys in any node, which have a name, you can do like this:

cv::FileStorage pfs(fileToRead, cv::FileStorage::READ);
cv::FileNode fn = pfs.root();
for (cv::FileNodeIterator fit = fn.begin(); fit != fn.end(); ++fit)
{
    cv::FileNode item = *fit;
    std::string somekey = item.name();
    std::cout << somekey << std::endl;
}

You can also do this recursively by iterating on the file node item

Answer 2

If the matrices are all mapped to the top level (as in your example), there is not a way to access them without knowing the name of their corresponding nodes. FileStorage::operator[] only takes string arguments specifying the name of the node.

A workaround could be to parse the YAML with another method to get the node names, and then access the FileStorage afterward.