the class attendance systems using face recognition and detection would work like this-
The duration of a class of 20 students(minimum) is usually 45 min on the average. The systems uses a camera as input device and takes snaps shots (or video)of students seated at predetermined time slots, detect their faces(students would not wear mask, caps etc and would not pose for their faces/pictures to be taken etc) and matches them against a known database of students that have registered for the class.
it then performs a recognition(multi) and updates the class register accordingly. The lecturer can then log into the system to know who attended his class etc.
.Students must be in class for a minimum duration of 70 % of the time(for instance, it should perform this face detection and recognition by taking real time pictures every 5 minutes and processing it).
You could make any addition as you deem fit and you could test it using any class room formation with a your friends seated and should be robust enough to give an accurate result atleast 90 percent of the time. An unauthorised student attending the class should be declared as unknown.
And as I advised earlier, you could use any of the known face recognition techniques( like adaboost with cascade detector, neural networks ,viola , gabor wavelet, open CV etc) and face detection techniques(subspace LDA etc).you might try simple techniques like eigenface, fisherface or PCA but the might not be robust enough.
NB-It could run either on Linux or Windows and their is flexibility as to technique. if you feel some threshold needs to be revisited to make this project implementable ,feel free to discuss your suggestion
14 freelancere byder i gennemsnit $815 på dette job
Hello I have worked on Face Recognition Algorithms before , I have Masters Degree in Computer Science and Statement of Accomplishment in Algorithms from Stanford . Thanks
This is a fairly large job to do properly. Need to optimize the pipeline from capturing the images to recognition and analysis. Please contact me for further information.