比如要生成一张图像(h=6, w=10)的xy坐标点，看下两者的实现方式：
两种方式的差异在于：
xs, ys = np.meshgrid(np.arange(w), np.arange(h))
xs, ys = torch.meshgrid([torch.arange(h), torch.arange(w)])
交换参数顺序，得到的结果一致。
具体实现如下：
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])
链接：https://www.jianshu.com/p/e1c8e51bc9b3
本文内容由网友自发贡献，版权归原作者所有，本站不承担相应法律责任。如您发现有涉嫌抄袭侵权的内容，请联系:hwhale#tublm.com(使用前将#替换为@)
