The data_set.xml file gathers in an organized way the path to your dataset.
At the first level of the hierarchy are the subjects, each one being identified through the <subject> [...] </subjects> tags.
Each subject has been seen to one or several visits, each one being identified through the <visit> [...] </visit> tags.
At each visit correspond a single age, given as a float value between the <age> [...] </age> tags. This age entry is optional for most of the Deformetrica applications.
At each visit one or several structures are observed, each one being identified through the <filename> [...] </filename> tags. The path to the corresponding shape data must be given between those tags.
No dataset: shoot and parallel transport
Those applications are direct computations, by opposition to estimating a statistical model from an input data-set. They do not require such a dataset.xml file, and are launched with the command: deformetrica model.xml [-p optimization_parameters.xml].
Single-visit dataset: register or parallel transport
From the examples/registration/landmark/3d/bundles_cortico example. Note that this example is a multi-object one, with two <filename> [...] </filename> tag within a single visit.
From the example examples/atlas/landmark/2d/skulls. Note that the age entry is not necessary, since only a single visit per subject is allowed in this cross-sectional setting. Note as well than the object_id tag associated to each filename must match the corresponding tag in the model.xml file, in the template section.