Ask Your Question

Time to train a cascaded classifier?

asked 2013-11-11 02:58:18 -0600

asaakius gravatar image

updated 2013-11-11 04:25:59 -0600

I have about 600 positive images and 1500 negative images from which I need to train a cascaded GentleBoost classifier (using very simple decision stumps as weak classifiers). About 2 thirds of all my images are 60x60, but the big ones are 480 x 320.

I'm going to need to train approximately 500 different classifiers and if possible would like to be doing parallel training.

I haven't started the training yet but I'm really worried as to how much time it might take, should I go for another implementation (like in MatLab)?

I'm on a workstation with 16GB RAM.


edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2013-11-11 06:04:26 -0600

For a 60x60 sized detection window the traincasade method needs about 2GB RAM. So, with 16 GB RAM you should be able to execute ~6 training methods in parallel, provided a good processor as well. You might want to compile the application on a 64-bit architecture to make better us of your RAM.

Training 500 classifiers will without a doubt take you a very long time either way. You don't have anything to gain (time-wise) by changing to MatLab

edit flag offensive delete link more


Thank goodness, I am already using a 64-bit architecture. 6 parallel classifiers training is somewhat a consolation - but approximately how long would the training be for one classifier? Let's say I only have 6 to train... how long approximately do you think the time range should be for a single training session?

asaakius gravatar imageasaakius ( 2013-11-11 07:35:39 -0600 )edit

I never trained multiple classifiers in parallel, so I really don't know. I can tell you that training a haar classifier with 600 positives and 1500 negatives shouldn't take a very long time, maybe half a day, possibly less, depending on your parameters. You should try it yourself and check if the time it takes is acceptable.

Pedro Batista gravatar imagePedro Batista ( 2013-11-11 07:54:16 -0600 )edit

Thanks very much! I just noticed something else, you understood my 60x60 resolution requirement as being related to my sliding detection window. That's correct. But is my image resolution reasonable? Is 480x320 reasonable or unnecessarily large? The final testing data won't have objects as big as 480x320, although the images and videos themselves maybe large.

asaakius gravatar imageasaakius ( 2013-11-11 08:23:43 -0600 )edit

Im not sure I understand what you are saying. If you are talking about performing detection in 480x320 images than that is no unreasonable at all. Many people use classifiers to perform detection in 640x480 images, and more.

Pedro Batista gravatar imagePedro Batista ( 2013-11-11 08:36:16 -0600 )edit

Oh yes I should have clarified - I was referring to the images used as training samples for the classifier, not for testing and final detection. The reason I am asking this is because in my understanding, having large images to train a classifier makes the training take longer. So I was wondering about 480x320 in terms of the image resolution for training samples - is that unnecessarily large, given that the objects I will be detecting in the end are unlikely to be that large.

asaakius gravatar imageasaakius ( 2013-11-11 09:00:54 -0600 )edit

Question Tools


Asked: 2013-11-11 02:58:18 -0600

Seen: 611 times

Last updated: Nov 11 '13