Probe samples may vary for a given protocol
Created by: laurentes
As previously discussed, I'm currently refactoring this database, making a better usage of the classes from models.py.
There are currently 5 protocols:
- 'all'
- 'expression'
- 'illumination'
- 'occlusion'
- 'occlusion_and_illumination'
However, it seems that depending on the parameters 'sessions', 'expressions', 'illuminations', 'occlusions', 'genders', the list of probe files will change (and the one of training files as well which is fine).
To my mind, this is a bad idea. We should define more specific protocols, for which enrollment and probe lists are fixed, whatever the parameters are. I propose to make these parameters world related. In this case, we would need to define more protocols to keep similar features. Which ones would you like?
The second question is what should be the default world data for each protocol.