So2sat processing steps

[AbdelghaniKABOT]

I trained a model on the So2sat dataset using TorchGeo and achieved excellent results on the testing split. However, when I tested it on new input data, the results were not as good. I think this may be due to differences in data processing.
I only performed normalization using the following function:

def normalize(array):
    """Normalizes numpy arrays into scale 0.0 - 1.0"""
    array_min, array_max = array.min(), array.max()
    return ((array - array_min)/(array_max - array_min))

Can you help me with this issue?"

[adamjstewart]

You should always use the same normalization as the pre-trained model. See here for the mean and std dev used to normalize images during training.