If I want to display one image from mnist dataset, I need to reshape it from (1,28,28) to (28,28) using the following code:

import tensorflow as tf
import matplotlib.pyplot as plt

mnist = tf.keras.datasets.mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0

sample = x_train[:1].reshape((28,28))
plt.imshow(sample, cmap="gray")
plt.show()

However, if I want to display multiple images within the same plot. I don't need to reshape them with the following code:

import tensorflow as tf
import matplotlib.pyplot as plt

mnist = tf.keras.datasets.mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0

plt.figure(figsize=(10,10))
for i in range(25):
  plt.subplot(5,5,i 1)
  plt.imshow(x_train[i])
plt.show()

Why reshape is not require in the second code?

CodePudding user response：

You wouldn't need reshape in the first one either if you selected the first image using x_train[0]. Accessing a specific index of the array removes the first element of the shape.

So if you have a numpy array of shape (100, 28, 28), and access x_train[0], you will get a shape of (28, 28). But if you access it as x_train[:1], you will still have three dimensions: (1, 28, 28).

It does this because you could also do x_train[:2] and take the first two images, so it needs a dimension to keep track of how many images you selected.