danqu130 / indian_pines_classification

A simple classifier using CNN with keras to classify Indian pines Hyperspectral Image

indian_pines_classification's Issues

about the window_Size

When running the code, I came across a question that when I increase the parameter window_Size even to a very large number, the classification accuracy still do not decrease. Instead, the accuracy increase with the increase of the window_Size, which is abnormal because the IndianPines dataset' spatial resolution is 20m. If the window_Size is too large, the feature extraction process will extract many noise features that do not belong to the center pixel.
So i wonder is there anyone else come across the same problem?
I would very like to have a discussion with you!