Using one-hot encoding to represent images

Neural networks expect the labels of classes in a dataset to be organized in a one-hot encoded manner: each row in the array contains zeros in all columns, except the column corresponding to a unique label, which is set to 1.

The fashion dataset contains three categories:

Shirts
Dresses
Shoes

In this exercise, you will create a one-hot encoding of a small sample of these labels.

This is a part of the course

“Image Modeling with Keras”

View Course

Exercise instructions

Initialize the ohe_labels variable to hold the one-hot encoded array.
Use np.where() to find the location of the category of the item in each iteration in categories.
Assign a 1 into the correct row/column combination in every iteration.

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# The number of image categories
n_categories = 3

# The unique values of categories in the data
categories = np.array(["shirt", "dress", "shoe"])

# Initialize ohe_labels as all zeros
ohe_labels = ____((len(labels), n_categories))

# Loop over the labels
for ii in range(len(labels)):
    # Find the location of this label in the categories variable
    jj = np.where(___)
    # Set the corresponding zero to one
    ohe_labels[____] = ____

Edit and Run Code