Face alignment detection with Fan2d106 only works well with a certain image types

[juliabeliaeva]

The model only return good results for images with big faces at the image center, and generally needs to be combined with a face detection method to detect the face bounding box. The following things need to be done:

• Add a desktop example of combining face detection and face alignment.
• Determine parameters for "bad" and "good" images.
• Add information to documentation.

[juliabeliaeva]

See #541