# tf.unique_with_counts(): Count the Number of Each Element in Tensor – TensorFlow Tutorial

By | July 8, 2021

In tensorflow, if we want to count the number of each value in a tensor, we can use tf.unique_with_counts() to implement. In this tutorial, we will use some examples to show you how to do.

## Syntax

tf.unique_with_counts() is defined as:

tf.unique_with_counts(
x, out_idx=tf.dtypes.int32, name=None
)

As to input $$x$$, it must be 1-D tensor.

This function will return three variables:

y, idx, count = tf.unique_with_counts(x)

Here

y: the unique value in tensor x

ids: the index id of value y in tensor x

count: the number of value y in x

Here is an example

# tensor 'x' is [1, 1, 2, 4, 4, 4, 7, 8, 8]
y, idx, count = unique_with_counts(x)
y ==> [1, 2, 4, 7, 8]
idx ==> [0, 0, 1, 2, 2, 2, 3, 4, 4]
count ==> [2, 1, 3, 1, 2]


More examples:

import tensorflow as tf
import numpy as np

x = np.array([1.0, 2.0, 3.0, 1.1, 1.0, 3.0])
x = tf.convert_to_tensor(x, dtype= tf.float32)

y, id, count = tf.unique_with_counts(x)

init = tf.global_variables_initializer()
init_local = tf.local_variables_initializer()
with tf.Session() as sess:
sess.run([init, init_local])
a =sess.run([y, id, count])
print(a)

Run this code, you will see these results:

[array([1. , 2. , 3. , 1.1], dtype=float32), array([0, 1, 2, 3, 0, 2], dtype=int32), array([2, 1, 2, 1], dtype=int32)]

From the result, we can find id and count is the int32.

## How about input x is not 1-D tensor?

Look at this example:

import tensorflow as tf
import numpy as np

x = np.array([1.0, 2.0, 3.0, 1.1, 1.0, 3.0])
x = tf.convert_to_tensor(x, dtype= tf.float32)
x = tf.reshape(x, [-1, 2])
y, id, count = tf.unique_with_counts(x)

init = tf.global_variables_initializer()
init_local = tf.local_variables_initializer()
with tf.Session() as sess:
sess.run([init, init_local])
a =sess.run([y, id, count])
print(a)

Here x will be 3*2 tensor. Running this code, you will get this error:

tensorflow.python.framework.errors_impl.InvalidArgumentError: unique expects a 1D vector.
[[{{node UniqueWithCounts}}]]

It means input x must be 1-D tensor.

This function will be used if you plan to compute center loss function.