YonghaoXu

Hi, the code is now updated with the `Keras 2.2.4` using the `Tensorflow` backend engine. This error is fixed now.

Hi, I think the `labels` here should be reshaped into a vector form (e.g., `labels = labels.reshape(145*145, 1)`).

Hi, you can use the `matplotlib` package to save the classification map. `import matplotlib.pyplot as plt` `...` `plt.imsave('clsmap.png',X_result)`

As mentioned in the paper, we use the same hyper-parameter setting for all three datasets. For experiments on the Indian Pines and KSC datasets, just replace ``dataID=1`` in ``SSUN.py`` to...

This error could be solved by changing the initialization of OASpectral_IP by ``OASpectral_IP = np.zeros((16+2,randtime))``, where ``16`` corresponds to the number of categories in the Indian Pines dataset.

The ``ProducerA`` vector contains the producer accuracy of the input data. Since there are ``16`` categories in the Indian Pines dataset, the ``ProducerA`` vector should also have ``16`` elements. Thus,...

You need to change the ``imageID`` in ``DrawResult`` func to generate the corresponding classification map. Please refer to the ``DrawResult`` func in ``HyperFunctions.py`` for details.

@TanmDL Thank you for your interest in our work. As described in the paper, the input of the MSCNN is a `w by w` neighbor region of each pixel in...

Hi, `s1s2` is the index to indicate the strategy utilized for band grouping. Strategy 1 ( `s1s2 = 1` case ) focuses on the local features and makes the signals...