The slicing method of three points [,...] in python

If you want to really understand slices, don’t have the concept of length, width and height. Instead, use the concept of dimensions.

$$PYTHOn-inthreethpoints\text{'}[,...]\text{'}Acutsheetsideof\; formula$$

The three points represent the operation on the lowest dimension, the so-called lowest dimension is the dimension X0 in the [Xn,Xn-1,...,X0] array

import torch
rgb = torch.randn(1, 3, 3, 4)
rgb

rgb.shape

rgb[...,1]

rgb[...,1].shape

rgb2 = torch.randn(3, 3, 4)

rgb2

rgb2[...,1]

rgb2[...,1].shape

---

import numpy
>>> a = numpy.array([[1,2,3,4,5],[6,7,8,9,10],[1,2,3,4,5],[6,7,8,9,10]])
>>> a
array([[ 1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10],
       [ 1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10]])
>>> a[...,2] #表示遍历每行，2表示索引为2的所在列
array([3, 8, 3, 8])
>>> a[...,:2]#表示遍历每行，:2表示索引为<2的0，1所在的列
array([[1, 2],
       [6, 7],
       [1, 2],
       [6, 7]])
>>> a[...,::2]#表示遍历每行，2表示步长，选取多索引为0，2，4所在的列
array([[ 1,  3,  5],
       [ 6,  8, 10],
       [ 1,  3,  5],
       [ 6,  8, 10]])
a[None,...]#相当于插入维度，也想当于reshape(a,[1,4,4])
array([[[[ 1,  2,  3,  4,  5],
         [ 6,  7,  8,  9, 10],
         [ 1,  2,  3,  4,  5],
         [ 6,  7,  8,  9, 10]]]])