Face Recognition with Large Training Set [closed]

asked 2015-07-11 19:14:45 -0500

lgaravaglia gravatar image

I'm working on a face recognition program based off of the OpenCV facerec demo. The training database that I would like to use contains 75,000 images, however trying to run the code with this setup results in my computer completely locking up.

Is there any way that the OpenCV implementation of face recognition can handle a large training set?

If not is there another free API for face recognition that would be able to handle a large training set?

edit retag flag offensive reopen merge delete

Closed for the following reason question is not relevant or outdated by sturkmen
close date 2020-10-15 11:07:04.121585


well, taking a pca of 75k images will take like a day or two. that's not a surprise, or is it ?

have you tried the lbph one ? you could partition your data, and let it run on several machines.

berak gravatar imageberak ( 2015-07-12 02:58:30 -0500 )edit

No, no surprise that it would take days to finish with PCA. It's just what I'm most familiar with at this point and I wanted to see what would actually happen if I tried to do it that way. I haven't tried the lbph method yet, I'm less familiar with that method and I made the assumption that it would behave in the same manner as PCA in regards to training on a data set that large. I'll take a more detailed look at it to see if it'll work for my case.

Partitioning the data is another method that I've put some thought into, unfortunately my lab is running low on resources right now and that may not be an option that I can pursue further.

lgaravaglia gravatar imagelgaravaglia ( 2015-07-12 13:01:04 -0500 )edit