Ask Your Question
0

Different results (foreground image) in MOG2 in 2.4.2~2.4.6

asked 2013-09-11 01:41:27 -0600

Vio gravatar image

Hi everyone, I used MOG2() to train/Learn 10 images and got the model. Use this model to detect current image in order to get foreground image. (When Training, Learning rate = 0.1; when Detecting, Learning rate = 0).

But I've discovered that it has different results (foreground images) in continuous 2 detection with the same MOG2 trained model. It has tiny difference between these 2 result images!!

But why ? I use the same MOG2 model to calculate it...

The 2 different results of the foreground image are displayed in the attached images. C:\fakepath\gmmfault.bmp (I am sure that the input image is the same when detecting) It looks like something wrong in the first 2 computation.

Greetings, Vio

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
0

answered 2013-09-11 02:07:37 -0600

Since 2.4.2 several updates have been applied to the algoritm that actually contained some bugs. A quick search in the developer forums gave me these results:

So I suggest grabbing the latest openCV version as it will always contain the most correct answer.

edit flag offensive delete link more

Question Tools

Stats

Asked: 2013-09-11 01:41:27 -0600

Seen: 1,381 times

Last updated: Sep 11 '13