For the one-hot encoding of the response variables, we use the following code:

# Labels
trainy <- c(0,0,0,1,1,1,2,2,2)
validy <- c(0,1,2)
testy <- c(0,0,1,1,2,2)

# One-hot encoding
trainLabels <- to_categorical(trainy)
validLabels <- to_categorical(validy)
testLabels <- to_categorical(testy)
trainLabels

OUTPUT
      [,1] [,2] [,3]
 [1,]    1    0    0
 [2,]    1    0    0
 [3,]    1    0    0
 [4,]    0    1    0
 [5,]    0    1    0
 [6,]    0    1    0
 [7,]    0    0    1
 [8,]    0    0    1
 [9,]    0    0    1

From the preceding code, we can see the following:

• We have stored target values for each image in trainy , validy, and testy, where 0, 1, and 2 indicate bicycle, car, and airplane images respectively.
• We carry out one-hot encoding of trainy , validy, and testy by using the to_categorical function. One-hot encoding here helps to convert a factor variable into a combination of zeros and ones.

Now we have the data in a format that can be used for developing a deep neural network classification model, and that is what we will do in the next section.