Data augmentation for images in TensorFlow - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to create a data augmentation layer that randomly flips images horizontally.

TensorFlow

data_augmentation = tf.keras.Sequential([
    tf.keras.layers.RandomFlip([1])
])

Drag options to blanks, or click blank then click option'

A'none'

B'vertical'

C'horizontal_and_vertical'

D'horizontal'

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'vertical' which flips images top to bottom instead of horizontally.

Using 'none' which disables flipping.

The RandomFlip layer with the argument 'horizontal' flips images left to right randomly, which is a common augmentation for image data.

2fill in blank

medium

Complete the code to add a random rotation of 20% to the data augmentation pipeline.

TensorFlow

data_augmentation = tf.keras.Sequential([
    tf.keras.layers.RandomRotation([1])
])

Drag options to blanks, or click blank then click option'

A0.2

B20

C2

D0.02

Attempts:

3 left

💡 Hint

Common Mistakes

Using 20 which is interpreted as 20 full rotations (invalid).

Using 2 which is more than 1 and invalid.

The RandomRotation layer expects a float between 0 and 1 representing the fraction of a full circle. 0.2 means 20% of 360 degrees, or about 72 degrees rotation.

3fill in blank

hard

Fix the error in the code to correctly apply random zoom augmentation.

TensorFlow

data_augmentation = tf.keras.Sequential([
    tf.keras.layers.RandomZoom(height_factor=[1])
])

Drag options to blanks, or click blank then click option'

A'0.1'

B10

C0.1

D-0.1

Attempts:

3 left

💡 Hint

Common Mistakes

Using a string instead of a float.

Using a negative value which is invalid.

Using a large integer which is out of range.

The height_factor expects a float between 0 and 1 representing the zoom range. 0.1 means zooming in or out by up to 10%.

4fill in blank

hard

Fill both blanks to create a data augmentation pipeline that randomly flips images vertically and applies random contrast adjustment.

TensorFlow

data_augmentation = tf.keras.Sequential([
    tf.keras.layers.RandomFlip([1]),
    tf.keras.layers.RandomContrast([2])
])

Drag options to blanks, or click blank then click option'

A'vertical'

B0.2

C0.5

D'horizontal'

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'horizontal' instead of 'vertical' for flipping.

Using contrast factor greater than 1 which is invalid.

RandomFlip('vertical') flips images top to bottom randomly. RandomContrast(0.2) adjusts contrast by up to 20%.

5fill in blank

hard

Fill all three blanks to create a data augmentation pipeline that randomly flips images horizontally, applies random rotation of 10%, and random zoom of 15%.

TensorFlow

data_augmentation = tf.keras.Sequential([
    tf.keras.layers.RandomFlip([1]),
    tf.keras.layers.RandomRotation([2]),
    tf.keras.layers.RandomZoom([3])
])

Drag options to blanks, or click blank then click option'

A'horizontal'

B0.1

C0.15

D'vertical'

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'vertical' flip instead of 'horizontal'.

Using rotation or zoom values greater than 1.

The pipeline flips images horizontally, rotates them by 10% of a full circle, and zooms by 15%. These augmentations help the model learn better by seeing varied images.