I'm trying to generate a synthetic sequence with MNIST images. Each image is flattened 784. When I select five of them, my data is shape (5,784). I want to concatenate 5 of them horizontally, that my final image has shape (28,5*28). How can I achieve that?
I tried it with np.reshape but the best I could achieve, was a vertical concatenation.
For demonstration, let's say we want to horizontally concatenate three images which are 4x4, stored flat as 16 elements:
a = np.arange(16)
b = np.arange(16,32)
c = np.arange(32,48)
images = np.array([a,b,c]) # 3x16
That's just to prepare the sample data. Now reshape and concatenate:
np.hstack(images.reshape(3,4,4))
The result is:
array([[ 0, 1, 2, 3, 16, 17, 18, 19, 32, 33, 34, 35],
[ 4, 5, 6, 7, 20, 21, 22, 23, 36, 37, 38, 39],
[ 8, 9, 10, 11, 24, 25, 26, 27, 40, 41, 42, 43],
[12, 13, 14, 15, 28, 29, 30, 31, 44, 45, 46, 47]])
In your original case, the expression would be:
np.hstack(stuff.reshape(5,28,28))
And indeed the shape is (28,140).
