np.meshgrid()函數(shù)常用于生成二維網(wǎng)格，比如圖像的坐標(biāo)點(diǎn)。
pytorch中也有一個(gè)類似的函數(shù)torch.meshgrid()，功能也類似，但是兩者的用法有區(qū)別，使用時(shí)需要注意（剛踩坑，因此記錄一下。。。）

比如我要生成一張圖像(h=6, w=10)的xy坐標(biāo)點(diǎn)，看下兩者的實(shí)現(xiàn)方式：

np.meshgrid()

>>> import numpy as np
>>> h = 6
>>> w = 10
>>> xs, ys = np.meshgrid(np.arange(w), np.arange(h))
>>> xs.shape
(6, 10)
>>> ys.shape
(6, 10)
>>> xs
array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
       [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
       [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
       [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
       [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
       [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]])
>>> ys
array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
       [2, 2, 2, 2, 2, 2, 2, 2, 2, 2],
       [3, 3, 3, 3, 3, 3, 3, 3, 3, 3],
       [4, 4, 4, 4, 4, 4, 4, 4, 4, 4],
       [5, 5, 5, 5, 5, 5, 5, 5, 5, 5]])
>>> xys = np.stack([xs, ys], axis=-1)
>>> xys.shape
(6, 10, 2)

torch.meshgrid()

>>> import torch
>>> h = 6
>>> w = 10
>>> ys,xs = torch.meshgrid(torch.arange(h), torch.arange(w))
>>> xs.shape
torch.Size([6, 10])
>>> ys.shape
torch.Size([6, 10])
>>> xs
tensor([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
        [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
        [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
        [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
        [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
        [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]])
>>> ys
tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
        [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
        [2, 2, 2, 2, 2, 2, 2, 2, 2, 2],
        [3, 3, 3, 3, 3, 3, 3, 3, 3, 3],
        [4, 4, 4, 4, 4, 4, 4, 4, 4, 4],
        [5, 5, 5, 5, 5, 5, 5, 5, 5, 5]])
>>> xys = torch.stack([xs, ys], dim=-1)
>>> xys.shape
torch.Size([6, 10, 2])

從python交互式窗口可以清晰的看出numpy和pytorch中meshgrid()函數(shù)的區(qū)別，就不用文字總結(jié)了，自己體會(huì)哈哈哈。