tf.where examples and its behavior

Before going into tf.where example, a proper way to implement equal comparison in tensorflow will be demonstrated since equal comparison will be used as the condition of the tf.where.

Equal Comparison in Tensorflow

import tensorflow as tf, os

a = tf.constant(\[\[1,1\],\[3,6\]\])
b= a==3

with tf.Session() as sess:
    output = sess.run(b)

    print(output)

output:

TypeError: Fetch argument False has invalid type <class 'bool'>, must be a string or Tensor. (Can not convert a bool into a Tensor or Operation.)

Process finished with exit code 1

to fix this, do not use python native operands such as == in the above. instead use comparing operations provided by tensorflow.

import tensorflow as tf, os

a = tf.constant(\[\[1,1\],\[3,6\]\])
b = tf.equal(a,3)

with tf.Session() as sess:
    output = sess.run(b)

    print(output)

output:

[[False False]
 [ True False]]

Now that we have verified how to properly use equal comparison in tensorflow, lets move on to tf.where

Examples of tf.where

import tensorflow as tf, os

a = tf.constant(\[\[1,1\],\[3,6\]\])
b = tf.where(tf.equal(a,3))

with tf.Session() as sess:
    output = sess.run(b)

    print(output)

output:

[[1 0]]

The coordinate of 3 in the sample tensor has been correctly identified. Then let’s try increasing the dimension of the target array.

import tensorflow as tf, os

a = tf.constant(\[\[\[1,1\],\[3,6\]\],\[\[7,8\],\[9,9\]\]\])
b = tf.where(tf.equal(a,3))

with tf.Session() as sess:
    output = sess.run(b)

    print(output)

output:

[[0 1 0]]

We can identify that the output of tf.where will be dependant on the dimension of the input tensor.

Then how would the output look like when there are multiple condition matches in the input tensor?

import tensorflow as tf, os

a = tf.constant(\[\[\[1,1\],\[3,6\]\],\[\[7,8\],\[3,3\]\]\])
b = tf.where(tf.equal(a,3))

with tf.Session() as sess:
    output = sess.run(b)

    print(output)

output:

[[0 1 0]
 [1 1 0]
 [1 1 1]]

We can identify that the coordinates of each of the multiple finding are returned in a list.