I want to reshape the MNIST digits dataset from (28, 28) to (32, 32). One way is to interpolate the data. I use custom Radial Basis Function for interpolation. How to do it??
Here is the RBF function
def RBF(x, c, s):
return np.exp(-np.sum((x-c)**2, axis=1)/(2*s**2))
where x is the actual value, c is the centre(assumed as mean) and s is standard deviation.
Here is how I load the data from tensorflow
import tensorflow as tf
data = tf.contrib.learn.datasets.mnist.load_mnist()
You can use, tf.image.resize_with_pad which resizes the image to a target width and height by keeping the aspect ration the same without distortion.
Below is the code for the same.
import tensorflow as tf
from matplotlib import pyplot as plt
(x_train, y_train), (x_test, y_test) =tf.keras.datasets.mnist.load_data(path='mnist.npz')
original_image = x_train[0].reshape(28,28,1)
resized_image = tf.image.resize_with_pad(original_image, 32,32)
plt.imshow(original_image.reshape(28,28))
Original Image:
Resized Image:
plt.imshow(resized_image.numpy().reshape(32,32))
You can see the scale range in both the images and the aspect ratio is also maintained.