如何使用 plt.imshow 和 torchvision.utils.make_grid 在 PyTorch 中生成和显示图像网格? [英] How can I generate and display a grid of images in PyTorch with plt.imshow and torchvision.utils.make_grid?

问题描述

我试图了解 torchvision 如何与 mathplotlib 交互以生成图像网格.生成图像并进行迭代显示很容易:

I am trying to understand how torchvision interacts with mathplotlib to produce a grid of images. It's easy to generate images and display them iteratively:

import torch
import torchvision
import matplotlib.pyplot as plt

w = torch.randn(10,3,640,640)
for i in range (0,10):
    z = w[i]
    plt.imshow(z.permute(1,2,0))
    plt.show()

但是，将这些图像显示在网格中似乎并不那么简单.

However, displaying these images in a grid does not seem to be as straightforward.

w = torch.randn(10,3,640,640)
grid = torchvision.utils.make_grid(w, nrow=5)
plt.imshow(grid)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-61-1601915e10f3> in <module>()
      1 w = torch.randn(10,3,640,640)
      2 grid = torchvision.utils.make_grid(w, nrow=5)
----> 3 plt.imshow(grid)

/anaconda3/lib/python3.6/site-packages/matplotlib/pyplot.py in imshow(X, cmap, norm, aspect, interpolation, alpha, vmin, vmax, origin, extent, shape, filternorm, filterrad, imlim, resample, url, hold, data, **kwargs)
   3203                         filternorm=filternorm, filterrad=filterrad,
   3204                         imlim=imlim, resample=resample, url=url, data=data,
-> 3205                         **kwargs)
   3206     finally:
   3207         ax._hold = washold

/anaconda3/lib/python3.6/site-packages/matplotlib/__init__.py in inner(ax, *args, **kwargs)
   1853                         "the Matplotlib list!)" % (label_namer, func.__name__),
   1854                         RuntimeWarning, stacklevel=2)
-> 1855             return func(ax, *args, **kwargs)
   1856 
   1857         inner.__doc__ = _add_data_doc(inner.__doc__,

/anaconda3/lib/python3.6/site-packages/matplotlib/axes/_axes.py in imshow(self, X, cmap, norm, aspect, interpolation, alpha, vmin, vmax, origin, extent, shape, filternorm, filterrad, imlim, resample, url, **kwargs)
   5485                               resample=resample, **kwargs)
   5486 
-> 5487         im.set_data(X)
   5488         im.set_alpha(alpha)
   5489         if im.get_clip_path() is None:

/anaconda3/lib/python3.6/site-packages/matplotlib/image.py in set_data(self, A)
    651         if not (self._A.ndim == 2
    652                 or self._A.ndim == 3 and self._A.shape[-1] in [3, 4]):
--> 653             raise TypeError("Invalid dimensions for image data")
    654 
    655         if self._A.ndim == 3:

TypeError: Invalid dimensions for image data

尽管 PyTorch 的文档表明 w 是正确的形状，但 Python 表示它不是.所以我试图置换我的张量的索引:

Even though PyTorch's documentation indicates that w is the correct shape, Python says that it isn't. So I tried to permute the indices of my tensor:

    w = torch.randn(10,3,640,640)
    grid = torchvision.utils.make_grid(w.permute(0,2,3,1), nrow=5)
    plt.imshow(grid)
---------------------------------------------------------------------------
    RuntimeError                              Traceback (most recent call last)
    <ipython-input-62-6f2dc6313e29> in <module>()
          1 w = torch.randn(10,3,640,640)
    ----> 2 grid = torchvision.utils.make_grid(w.permute(0,2,3,1), nrow=5)
          3 plt.imshow(grid)

    /anaconda3/lib/python3.6/site-packages/torchvision-0.2.1-py3.6.egg/torchvision/utils.py in make_grid(tensor, nrow, padding, normalize, range, scale_each, pad_value)
         83             grid.narrow(1, y * height + padding, height - padding)\
         84                 .narrow(2, x * width + padding, width - padding)\
    ---> 85                 .copy_(tensor[k])
         86             k = k + 1
         87     return grid

    RuntimeError: The expanded size of the tensor (3) must match the existing size (640) at non-singleton dimension 0

这是怎么回事?如何将一堆随机生成的图像放入网格中并显示它们?

What's happening here? How can I place a bunch of randomly generated images into a grid and display them?

推荐答案

您的代码中有一个小错误.例如，以下代码可以正常工作:

There's a small mistake in your code. For example, the below code works fine:

In [107]: import torchvision

# sample input (10 RGB images containing just Gaussian Noise)
In [108]: batch_tensor = torch.randn(*(10, 3, 256, 256))   # (N, C, H, W)

# make grid (2 rows and 5 columns) to display our 10 images
In [109]: grid_img = torchvision.utils.make_grid(batch_tensor, nrow=5)

# check shape
In [110]: grid_img.shape
Out[110]: torch.Size([3, 518, 1292])

# reshape and plot (because MPL needs channel as the last dimension)
In [111]: plt.imshow(grid_img.permute(1, 2, 0))
Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
Out[111]: <matplotlib.image.AxesImage at 0x7f62081ef080>

其输出显示为:

这篇关于如何使用 plt.imshow 和 torchvision.utils.make_grid 在 PyTorch 中生成和显示图像网格?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！