Data augmentation in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to apply a random horizontal flip to an image tensor using PyTorch transforms.

PyTorch

transform = torchvision.transforms.Compose([torchvision.transforms.RandomHorizontalFlip(p=[1])])

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

A2

B1.5

C0.5

D-0.5

Attempts:

3 left

💡 Hint

Common Mistakes

Using a probability greater than 1 or less than 0.

Forgetting to set the probability parameter.

The RandomHorizontalFlip transform takes a probability p between 0 and 1 to flip the image horizontally. 0.5 means 50% chance.

2fill in blank

medium

Complete the code to normalize an image tensor with mean 0.5 and std 0.5 using PyTorch transforms.

PyTorch

transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor(), torchvision.transforms.Normalize(mean=[[1]], std=[0.5])])

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

A0.0

B-0.5

C1.0

D0.5

Attempts:

3 left

💡 Hint

Common Mistakes

Using mean 0 or 1 which may not center the data properly.

Confusing mean and std values.

The Normalize transform requires mean and std values for each channel. Here, mean is set to 0.5 to center the data.

3fill in blank

hard

Fix the error in the code to apply random rotation of up to 30 degrees to an image tensor.

PyTorch

transform = torchvision.transforms.Compose([torchvision.transforms.RandomRotation(degrees=[1])])

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

A[30]

B(-30, 30)

C30

D30.0

Attempts:

3 left

💡 Hint

Common Mistakes

Passing a single integer instead of a range tuple.

Passing a list instead of a tuple.

The RandomRotation transform expects a range tuple for degrees to rotate randomly between. Passing a tuple like (-30, 30) is correct.

4fill in blank

hard

Fill both blanks to create a transform that converts an image to tensor and then applies a random vertical flip with 30% chance.

PyTorch

transform = torchvision.transforms.Compose([[1](), torchvision.transforms.RandomVerticalFlip(p=[2])])

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

AToTensor

BRandomHorizontalFlip

C0.3

D0.7

Attempts:

3 left

💡 Hint

Common Mistakes

Using horizontal flip instead of vertical flip.

Setting probability to 0.7 instead of 0.3.

First, images must be converted to tensors using ToTensor(). Then, RandomVerticalFlip with probability 0.3 flips images vertically 30% of the time.

5fill in blank

hard

Fill all three blanks to create a transform pipeline that resizes images to 128x128, converts to tensor, and normalizes with mean 0.5 and std 0.5.

PyTorch

transform = torchvision.transforms.Compose([torchvision.transforms.[1]((128, 128)), torchvision.transforms.[2](), torchvision.transforms.Normalize(mean=[[3]], std=[0.5])])

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

AResize

BToTensor

C0.5

DCenterCrop

Attempts:

3 left

💡 Hint

Common Mistakes

Using CenterCrop instead of Resize.

Forgetting to convert images to tensors before normalization.

The Resize transform changes image size, ToTensor converts images to tensors, and Normalize centers and scales pixel values.