PDF version of this entire document

FRGC Experiments

As template experiments with data that was collected from distinct subjects are needed (in order, for example, to perform recognition tests and use a classifier for ROC curves), experimental protocols from FRGC were sought, finding a suitable training set and target/query/matching sets.

This process varies from learning some of the metadata and experiment scripts/SGML/markup/data to looking at some sample data in 3-D, which is non-trivial to browse/navigate without some special/ised visualisation tools. For FRGC 2.0, the master file in

FRGC-2.0-dist/BEE_DIST/FRGC2.0/metadata/FRGC_2.0_Metadata.xml

contains a suitable index and experiments from previous years can also be run using the older protocols at

FRGC-2.0-dist/BEE_DIST/FRGC1.0

Early tests with the data (see screenshots of the development and experimentation framework in Figure , as well as figures , , and ) suggest that we need to make some new binary masks for residuals.

Figure: Example 3-D representation of an arbitrary face image from FRGC 2.0

Figure: Example of alignment and cropping of the rigid parts of another face (mind axes scale) with the result shown at the top surface and right image inside the GUI

Figure: Example of alignment and cropping of the rigid parts of another face with the result shown at the top surface and right image inside the GUI

Figure: Example of alignment and cropping of the rigid parts of another face there there is some noise and hole that pose a challenge

``The right to be heard does not automatically include the right to be taken seriously.''
-Hubert Humphrey.