kkgadiraju/sat-classification-using-cnn

Classifying SAT-6 data using a CNN

sat-classification-using-cnn's Issues

CSV to Image processing weirdness - DeepSat Sat6

Hi there,

finally found something to read in the DeepSat Sat6 data set ( http://csc.lsu.edu/~saikat/deepsat/) provided by @kkgadiraju (SAT-Classification-Using-CNN/Preprocessing-SAT6.ipynb). However, I'm, not sure, whether the image output is correct.

Here is @kkgadiraju respective code:

data = sio.loadmat('/home/gadiraju/data/bl-slums/sat-6-full.mat')
train_X = data['train_x'].transpose()
train_Y = data['train_y'].transpose()
test_X = data['test_x'].transpose()
test_Y = data['test_y'].transpose()
annotations = data['annotations']

print train_X.shape, train_Y.shape, test_X.shape, test_Y.shape, annotations

train_X = train_X.reshape(train_X.shape[0], 28,28,4)
test_X = test_X.reshape(test_X.shape[0],28,28,4)full 

print train_X.shape, test_X.shape

When I try to visualize it with

ix =100001#Type a number between 0 and 323,999 inclusive
imshow(np.squeeze(train_X[ix,:,:,0:3]).astype(int)) #Only seeing the RGB channels
plt.show()
#Tells what the image is
if train_Y[ix,0] == 1:
    print ('Barren Land')
elif train_Y[ix,1] == 1:
    print ('Trees')
elif train_Y[ix,2] == 1:
    print ('Grassland')
elif train_Y[ix,3] == 1:
    print ('Roads')
elif train_Y[ix,4] == 1:
    print ('Buildings')
elif train_Y[ix,5] == 1:
    print ('Water Body')
else:
    print ('Other')

the resulting image is somewhat odd and patterned estranged, not showing actually what is labelled (in my opinion). Anybody there who knows the issue? Guess its just a minor one, but I somewhat do not notice, what it is.