how can face recognize with one sample per object
to implement fisher and eigenface algoritm need to have more than 2 sample for one image in database. Which algorithm is suitable for a sample؟
If you want to do a recognition of one object in another image, then you shall start with this. If you have more than one image per class than you need some kind of classifier/recognizer; but I have not seen a recognizer based on one sample per class.
just to scare you a bit: http://vis-www.cs.umass.edu/lfw/resul... ;)
we have only 1 examples of each face. may create multiple copies of a photo Which are different in terms of brightness and angle,(3D Morphabe) , for use fisherface algorithm.
That will not work ... generating artificial data for making a classifier that should then work in actual applications is the worst possible approach I see people doing in academic research!
@jal, please realize, that you picked a real hard problem , and that you won't get away without investing a significant amount of work here.
e.g. you need lighting normalization anyway, so artificially changing the illumination won't give any win. even if you flip() the image, you do not get a real new one, though it is done often to augment the input data, the gain is very low.
with only one img per person, you just can't do a identification (one img against a database), you probably have to do a verification (same, not-same test) , which is exponentially harder to do. please look again at the lfw link above, to roughly estimate the cost of your project..