TensorFlow - How to add padding to a tensor

TensorFlow is open-source Python library designed by Google to develop Machine Learning models and deep learning  neural networks.

Padding means adding values before and after Tensor values.

Method Used:

• tf.pad: This method accepts input tensor and padding tensor with other optional arguments and returns a Tensor with added padding and same type as input Tensor. Padding tensor is a Tensor with shape(n, 2).

Example 1: This example uses constant padding mode i.e. value at all the padded indices will be constant.

# importing the library
import tensorflow as tf

# Initializing the Input
input = tf.constant([[1, 2], [3, 4]])
padding = tf.constant([[2, 2], [2, 2]])

# Printing the Input
print("Input: ", input)
print("Padding: ", padding)

# Generating padded Tensor
res = tf.pad(input, padding, mode ='CONSTANT')

# Printing the resulting Tensors
print("Res: ", res )

Output:

Input:  tf.Tensor(
[[1 2]
 [3 4]], shape=(2, 2), dtype=int32)
Padding:  tf.Tensor(
[[2 2]
 [2 2]], shape=(2, 2), dtype=int32)
Res:  tf.Tensor(
[[0 0 0 0 0 0]
 [0 0 0 0 0 0]
 [0 0 1 2 0 0]
 [0 0 3 4 0 0]
 [0 0 0 0 0 0]
 [0 0 0 0 0 0]], shape=(6, 6), dtype=int32)

Example 2: This example uses REFLECT padding mode. For this mode to work paddings[D, 0] and paddings[D, 1] must be less than or equal to tensor.dim_size(D) - 1.

# importing the library
import tensorflow as tf

# Initializing the Input
input = tf.constant([[1, 2, 5], [3, 4, 6]])
padding = tf.constant([[1, 1], [2, 2]])

# Printing the Input
print("Input: ", input)
print("Padding: ", padding)

# Generating padded Tensor
res = tf.pad(input, padding, mode ='REFLECT')

# Printing the resulting Tensors
print("Res: ", res )

Output:

Input:  tf.Tensor(
[[1 2 5]
 [3 4 6]], shape=(2, 3), dtype=int32)
Padding:  tf.Tensor(
[[1 1]
 [2 2]], shape=(2, 2), dtype=int32)
Res:  tf.Tensor(
[[6 4 3 4 6 4 3]
 [5 2 1 2 5 2 1]
 [6 4 3 4 6 4 3]
 [5 2 1 2 5 2 1]], shape=(4, 7), dtype=int32)

aman neekhara

Follow

Improve

Article Tags :

• Python
• Tensorflow
• Python-Tensorflow

Practice Tags :

• python