向 keras 序列模型添加调整大小层 [英] Add a resizing layer to a keras sequential model

问题描述

如何向

model = Sequential()

使用

model.add(...)

要将图像的大小从 (160, 320, 3) 调整为 (224,224,3) 吗?

To resize an image from shape (160, 320, 3) to (224,224,3) ?

推荐答案

通常你会使用 Reshape 图层:

Normally you would use the Reshape layer for this:

model.add(Reshape((224,224,3), input_shape=(160,320,3))

但是由于您的目标维度不允许包含来自输​​入维度 (224*224 !=160*320) 的所有数据，这将不起作用.如果元素数量不变，您只能使用 Reshape.

but since your target dimensions don't allow to hold all the data from the input dimensions (224*224 != 160*320), this won't work. You can only use Reshape if the number of elements does not change.

如果您不介意丢失图像中的某些数据，则可以指定自己的有损重塑:

If you are fine with losing some data in your image, you can specify your own lossy reshape:

model.add(Reshape(-1,3), input_shape=(160,320,3))
model.add(Lambda(lambda x: x[:50176])) # throw away some, so that #data = 224^2
model.add(Reshape(224,224,3))

也就是说，这些转换通常是在将数据应用到模型之前完成的，因为如果在每个训练步骤中都进行，这实际上是在浪费计算时间.

That said, often these transforms are done before applying the data to the model because this is essentially wasted computation time if done in every training step.

这篇关于向 keras 序列模型添加调整大小层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！