FileStorage for OpenCV Python API

You can use PyYAML to parse the YAML file.

Since PyYAML doesn't understand OpenCV data types, you need to specify a constructor for each OpenCV data type that you are trying to load. For example:

import yaml
def opencv_matrix(loader, node):
    mapping = loader.construct_mapping(node, deep=True)
    mat = np.array(mapping["data"])
    mat.resize(mapping["rows"], mapping["cols"])
    return mat
yaml.add_constructor(u"tag:yaml.org,2002:opencv-matrix", opencv_matrix)

Once you've done that, loading the yaml file is simple:

with open(file_name) as fin:
    result = yaml.load(fin.read())

Result will be a dict, where the keys are the names of whatever you saved in the YAML.

Using the FileStorage functions available in OpenCV 3.2, I've used this with success:

import cv2
fs = cv2.FileStorage("calibration.xml", cv2.FILE_STORAGE_READ)
fn = fs.getNode("Camera_Matrix")
print (fn.mat())

In addition to @misha's response, OpenCV YAML's are somewhat incompatible with Python.

Few reasons for incompatibility are:

1. Yaml created by OpenCV doesn't have a space after ":". Whereas Python requires it. [Ex: It should be a: 2, and not a:2 for Python]
2. First line of YAML file created by OpenCV is wrong. Either convert "%YAML:1.0" to "%YAML 1.0". Or skip the first line while reading.

The following function takes care of providing that:

import yaml
import re
def readYAMLFile(fileName):
    ret = {}
    skip_lines=1    # Skip the first line which says "%YAML:1.0". Or replace it with "%YAML 1.0"
    with open(scoreFileName) as fin:
        for i in range(skip_lines):
            fin.readline()
        yamlFileOut = fin.read()
        myRe = re.compile(r":([^ ])")   # Add space after ":", if it doesn't exist. Python yaml requirement
        yamlFileOut = myRe.sub(r': \1', yamlFileOut)
        ret = yaml.load(yamlFileOut)
    return ret

outDict = readYAMLFile("file.yaml")

NOTE: Above response is applicable only for yaml's. XML's have their own share of problems, something I haven't explored completely.

I wrote a small snippet to read and write FileStorage-compatible YAMLs in Python:

# A yaml constructor is for loading from a yaml node.
# This is taken from @misha 's answer: http://stackoverflow.com/a/15942429
def opencv_matrix_constructor(loader, node):
    mapping = loader.construct_mapping(node, deep=True)
    mat = np.array(mapping["data"])
    mat.resize(mapping["rows"], mapping["cols"])
    return mat
yaml.add_constructor(u"tag:yaml.org,2002:opencv-matrix", opencv_matrix_constructor)

# A yaml representer is for dumping structs into a yaml node.
# So for an opencv_matrix type (to be compatible with c++'s FileStorage) we save the rows, cols, type and flattened-data
def opencv_matrix_representer(dumper, mat):
    mapping = {'rows': mat.shape[0], 'cols': mat.shape[1], 'dt': 'd', 'data': mat.reshape(-1).tolist()}
    return dumper.represent_mapping(u"tag:yaml.org,2002:opencv-matrix", mapping)
yaml.add_representer(np.ndarray, opencv_matrix_representer)

#examples 

with open('output.yaml', 'w') as f:
    yaml.dump({"a matrix": np.zeros((10,10)), "another_one": np.zeros((2,4))}, f)

with open('output.yaml', 'r') as f:
    print yaml.load(f)

To improve on the previous answer by @Roy_Shilkrot I added support for numpy vectors as well as matrices:

# A yaml constructor is for loading from a yaml node.
# This is taken from @misha 's answer: http://stackoverflow.com/a/15942429
def opencv_matrix_constructor(loader, node):
    mapping = loader.construct_mapping(node, deep=True)
    mat = np.array(mapping["data"])
    if mapping["cols"] > 1:
        mat.resize(mapping["rows"], mapping["cols"])
    else:
        mat.resize(mapping["rows"], )
    return mat
yaml.add_constructor(u"tag:yaml.org,2002:opencv-matrix", opencv_matrix_constructor)


# A yaml representer is for dumping structs into a yaml node.
# So for an opencv_matrix type (to be compatible with c++'s FileStorage) we save the rows, cols, type and flattened-data
def opencv_matrix_representer(dumper, mat):
    if mat.ndim > 1:
        mapping = {'rows': mat.shape[0], 'cols': mat.shape[1], 'dt': 'd', 'data': mat.reshape(-1).tolist()}
    else:
        mapping = {'rows': mat.shape[0], 'cols': 1, 'dt': 'd', 'data': mat.tolist()}
    return dumper.represent_mapping(u"tag:yaml.org,2002:opencv-matrix", mapping)
yaml.add_representer(np.ndarray, opencv_matrix_representer)

Example:

with open('output.yaml', 'w') as f:
    yaml.dump({"a matrix": np.zeros((10,10)), "another_one": np.zeros((5,))}, f)

with open('output.yaml', 'r') as f:
    print yaml.load(f)

Output:

a matrix: !!opencv-matrix
  cols: 10
  data: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
    0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
    0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
    0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
    0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
    0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
    0.0, 0.0, 0.0, 0.0, 0.0]
  dt: d
  rows: 10
another_one: !!opencv-matrix
  cols: 1
  data: [0.0, 0.0, 0.0, 0.0, 0.0]
  dt: d
  rows: 5

Though I could not control the order of rows, cols, dt, data.

pip install opencv-contrib-python for video support to install specific version use pip install opencv-contrib-python