Numpy 将形状从 (3, 512, 660, 4) 更改为 (3,2048,660,1) [英] Numpy change shape from (3, 512, 660, 4) to (3,2048,660,1)

问题描述

我正在研究 tensorflow，我在一个形状为 (3, 512, 660, 4) 的 numpy 数组中有一些汽车图像.

I am working on tensorflow and I am having some images of cars in an numpy array with shape (3, 512, 660, 4).

这里，3对应一个汽车索引，512*660是一个图片尺寸，4对应一个车标的不同边车.

In this, 3 corresponds to a car index, 512*660 is an image size and 4 corresponds to the different sides of a car.

即(1, 512, 660, 1)对应Car1 - 正面图，(1, 512, 660, 2)对应Car1 -左侧图像等.

That is, (1, 512, 660, 1) corresponds to Car1 - front side image, (1, 512, 660, 2) corresponds to Car1 - Left side image and so on.

现在，我想将汽车的所有图像合并为一个图像(2048*660).也就是说，我想将 (3, 512, 660, 4) 重塑为 (3, 2048, 660, 1).

Now, I want to concat all the images of a car into one image (2048*660). That is, I want to reshape (3, 512, 660, 4) to (3, 2048, 660, 1).

有人可以帮我吗?

我尝试了 reshape 函数，但它实际上是重叠图像而不是连接它.

I tried reshape function but it actually overlaps images rather than concatenating it.

推荐答案

我们可以置换坐标轴，将最后一个坐标轴作为新的第三个坐标轴向前推进并重塑.置换轴可以用 np.swapaxes 或 np.transpose 或 np.rollaxis 处理，给我们三个解决方案，就像这样 -

We could permute axes to push the last axis up front as the new third axis and reshape. Permuting axes could be handled with np.swapaxes or np.transpose or np.rollaxis, giving us three solutions, like so -

a.swapaxes(2,3).reshape(3,2048,660,1)
a.transpose(0,1,3,2).reshape(3,2048,660,1)
np.rollaxis(a,3,2).reshape(3,2048,660,1)

如果你想在前面有边索引，相应地转置它 -

If you wanted to have sides-index at the front, transpose it accordingly -

a.transpose(0,3,1,2).reshape(3,2048,660,1)

这篇关于Numpy 将形状从 (3, 512, 660, 4) 更改为 (3,2048,660,1)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！