NxN 矩阵的筛选描述符列表

Question

I have a lot of sift descriptor from a dense-sift algorithm. Its an Array of N SIFT descriptor. One descriptor looks like this one:

[14.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 2.0, 0.0, 0.0, 0.0, 16.0, 0.0, ...]

y Goal is to transform the descriptors to an NxN Matrix so i can use them like an image feature input for an Machine Learning Struktur. So in Short: I want to transform many 1d arrays to an 2d Array taking into account that the data points are split in the correct order.

Is there a way to work with np.vstack or is there a other approach?

Answer 1

You can use stack and reshape and specify the desired axis like the following:

import numpy as np

arr1 = np.ones((1, 128))
arr1 = arr1.reshape((4,4,8))

arr2 = np.ones((1, 128))
arr2 = arr2.reshape((4,4,8))


print(np.stack((arr1, arr2), axis=3).shape)
# (4, 4, 8, 2)