GSOC project progress

Vaibhav Sharma's picture

Today I worked on integrating the LDS function with the face detection system. For that i worked on the initial step of the testing procedure where i label top 10 most confident vectors of each class. I extended that function for multiple classes as initially i was testing MAD on only two classes with videos.Also I think for multiple classes i would have to decrease the number of confident vectors i take because 10 vectors of each class would make the testing procedure supervised as in semi supervised labelled data is less.