learnopencv
learnopencv copied to clipboard
Training a Custom Classifier on a Custom Dataset for BRISQUE
Hi & Thanks for the awesome codebase
The BRISQUE's trainModel()
function has a couple of internal variables that I can't get a right grasp on to train my own custom SVM classifier:
-
Where do the values for "livedbfiles/dmos.txt" Come from? they are used as
differential mean opinion score (DMOS)
scores for each distorted image's quality. If I want to train my own SVM on a Binary (Good/Bad) dataset, should I just give0
to the "Good" images and100
to the Bad ones? -
If I want to train my own custom SVM classifier, Do I have to have a dataset like the "LIVE Image Quality Assessment Database" dataset where we have
original
anddistorted
versions of theoriginal
image or can I have a dataset full of different good & bad quality images and expect the model to learn a proper mapping for them?
Thanks again for the awesome work.