F.grid_sample

torch.nn.functional.grid_sample()函数的参数grid，表示的是范围为[-1, 1]坐标系下的(x, y, z)，坐标与数组的对应关系是：

x -> w, y -> h, z -> d，测试代码如下：

1. import numpy as np

2. from torch.nn import functional as F

3. import torch

4.  

5. if __name__ == '__main__':

6. d, h, w = 8, 10, 12

7. input = torch.zeros((2, 1, 8, 10, 12), dtype=torch.float32)

8. input[:, 0, 2, 3, 4] = 1

9. grid = torch.zeros((2, 1, 1, 1, 3), dtype=torch.float32)

10. x, y, z = 4, 3, 2 # 对应input的w, h, d

11. # rescale to [-1, 1]

12. x = 2. * x / (w - 1) - 1.

13. y = 2. * y / (h - 1) - 1.

14. z = 2. * z / (d - 1) - 1.

15. grid[0, 0, 0, 0, :] = torch.from_numpy(np.array([x, y, z]).astype(np.float32))

16. grid[1, 0, 0, 0, :] = torch.from_numpy(np.array([x, y, z]).astype(np.float32))

17. out = F.grid_sample(input, grid, mode='nearest')

18. print(out)###torch.Size([2, 1, 1, 1, 1])

可以看到输出为：

tensor([[[[[1.]]]],

        [[[[1.]]]]])