CODEWITHSUNDEEP
pythonarraysnumpy
Richard X
Richard X

Reputation: 1134

Reshape 4D batch of tiles to 3D Image in Numpy

Essentially, I have an array of images in the shape of [36, 107, 107, 3]. I want to reshape this batch of images so that the resulting array is one image of shape [642, 642, 3]. Everything that I have tried so far has resulted in the image being distorted. Any help is appreciated.

Upvotes: 0

Views: 165

Answers (1)

zephyrus
zephyrus

Reputation: 1266

Note, this will not be efficient, but it can be accomplished with some iteration:

ims = np.ones(36, 107, 107, 3)

im_per_side = np.sqrt(ims.shape[0]).astype(int)

# Reshape so we have real axes with desired orientation of images.
ims = np.reshape(ims, [im_per_side, im_per_side, 107, 107, 3]

length_slices = []
for l in range(im_per_side):
  height_slice = []
  for h in range(im_per_side):
    height_slice.append(ims[l, h])
  length_slices.apppend(np.concatenate(im_layer, 1))

# Finally concatenate along the `length` axis.
final_ims = np.concatenate(length_slices, 0)

Upvotes: 1

Related Questions