1 | initial version |
if you try to find out, what LDA does , you'll see that it
" attempts to model the difference between the classes of data"
obviously, to find a difference, you need more than 1 class (different persons)
to make the PCA (the previous step) work, you will also need more images per person, like 5-30.
last, just saying: opencv's face recognition classes do identification (which out of N in the database). if you wanted to find out, if 2 images show the same person, that's verification, something completely different. also, authentification (let only 1 person in). make sure to use the right tool for your job.