A multi-algorithmic face recognition framework for automatic attendance-taking

The recent economic downturn has changed the way universities receive funding, the new models place emphasis on retention and completion of classes, in addition to enrollment. Given that there is a strong correlation between attendance and class completion, faculty has been directed to take attendance into account for grades. At the same time budget constraints have increased the number of students per class, making attendance taking in a fast and efficient way an interesting problem with possible technical solutions.;To solve the attendance-taking problem a framework was developed. The framework consisted of a set of face recognition algorithms that were base lined against a well-known data set. Data was captured in a classroom setting and processed through the framework. The results were analyzed in terms of memory, speed and accuracy.;The results showed that some algorithms were faster (under 100 microseconds per image) and memory efficient (less than 5 Megabytes for training and testing), as well as reasonably accurate (above 80% recognition rates), while other missed the mark entirely, requiring too much tuning and orders of magnitude more memory.;It is concluded that the solution is feasible, but more data must be captured per class session in order to create uniform distributions, which are required for the algorithms to work optimally.