Modeling dependencies in COVAMOF

1
Modeling dependencies in COVAMOF

• Marco Sinnema, Sybren Deelstra,
• Jos Nijhuis, Jan Bosch
• University of Groningen

2
Dacolian Intrada Software Product Family
Monitoring
Parking
Access Control
3
Intrada Variabilty

• Countries
• Recognition Rate
• Error Rate
• Performance
• Memory Consumption
• Etc.

4
Intrada product family
ALPR
5
COVAMOF Variability View
6
Dependencies in COVAMOF

• Simple Dependencies
• Statically analyzable
• Function valid

• Complex Dependencies
• Dynamically analyzable
• Aspect (Real A)
• Valid range
• VP Associations
• Test
• Strategy
• Tested

Example at Dacolian Memory usage Error rate
7
Example from Intrada
8
Example

• (DEP_ID MEMORY_USAGE),
• (DESCRIPTION The maximum consumption in
  kilobytes by the ALPR library),
• (ASPECT memory_consumption),
• (VALIDRANGE ?A ? if A lt 64000 then true else
  false),
• (TEST Compile the product and automatically
  recognize 10000 images.),
• (STRATEGY Based on the required countries,
  first select the appropriate Matchers.
• Then find a suitable set of Neural Networks for
  these countries that fit in the
• remaining amount of memory. ),
• (ASSOCIATIONS
• (VP_ID VP3),
• (ASSOCTYPE DIRECTIONAL),
• (DIRECTION FB1,FB4,FB1,FB2,FB4,FB1,FB
  2,FB3,FB4 ),
• (VP_ID VP4),
• (ASSOCTYPE PREDICTABLE),
• (PREDICTION ?A ? A selected(Omni)3500
  selected(Digit)1700
• selected(Doubt)3200 selected(Alpha)8
  00 )
• )

9
Questions?

• More information
• http//segroup.cs.rug.nl/covamof