Applying the classifer with Trainable weka segmentation on part of image i.e. ROI give different results

Firstly, I trained a classifer model with Trainable weka segmentation. Then i applied the same classifier on the image it resulted in classified image. Secondly, i set the ROI and cropped that image to apply the same classifier. The result contains more regions classified as my object(false classification) as compared to the first classified image. Why is it so?
Just because of cropping an image why do classification results differ?

This might happen because the features extracted from the larger image contain more context information than the ROI version of it. In other words, the decisions made by the classifier were taking into account information that is no longer there when you crop.

1 Like

Thanks a lot for your reply

1 Like