Data augmentation with transforms in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the torchvision transforms module.

PyTorch

from torchvision import [1]

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

Autils

Btransforms

Cdatasets

Dmodels

Attempts:

3 left

💡 Hint

Common Mistakes

Importing datasets instead of transforms

Importing models which is unrelated to data augmentation

The transforms module in torchvision provides common image transformations used for data augmentation.

2fill in blank

medium

Complete the code to create a transform that randomly flips images horizontally.

PyTorch

transform = transforms.[1](p=0.5)

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

AColorJitter

BRandomRotation

CRandomHorizontalFlip

DRandomCrop

Attempts:

3 left

💡 Hint

Common Mistakes

Using RandomRotation which rotates images instead of flipping

Using ColorJitter which changes colors, not flips

RandomHorizontalFlip randomly flips the image horizontally with a given probability p.

3fill in blank

hard

Fix the error in the transform pipeline to convert images to tensor after resizing.

PyTorch

transform = transforms.Compose([
    transforms.Resize((128, 128)),
    transforms.[1](),
])

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

AToTensor

BToPILImage

CNormalize

DRandomCrop

Attempts:

3 left

💡 Hint

Common Mistakes

Using ToPILImage which converts tensor back to image

Using Normalize without converting to tensor first

ToTensor converts a PIL image or numpy array to a PyTorch tensor, which is needed after resizing.

4fill in blank

hard

Fill both blanks to create a transform pipeline that randomly crops and then normalizes images.

PyTorch

transform = transforms.Compose([
    transforms.[1](size=100),
    transforms.ToTensor(),
    transforms.[2](mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
])

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

ARandomCrop

BRandomHorizontalFlip

CNormalize

DResize

Attempts:

3 left

💡 Hint

Common Mistakes

Using Resize instead of RandomCrop for cropping

Using RandomHorizontalFlip instead of Normalize for normalization

RandomCrop crops the image randomly to the given size. Normalize adjusts pixel values to have the specified mean and standard deviation.

5fill in blank

hard

Fill all three blanks to create a transform pipeline that resizes, converts to tensor, and applies random rotation.

PyTorch

transform = transforms.Compose([
    transforms.[1]((64, 64)),
    transforms.[2](),
    transforms.[3](degrees=30)
])

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

AResize

BToTensor

CRandomRotation

DCenterCrop

Attempts:

3 left

💡 Hint

Common Mistakes

Mixing order of transforms causing errors

Using CenterCrop instead of Resize for resizing

The pipeline first Resizes images to 64x64 pixels, then converts them to tensors with ToTensor, and finally applies a random rotation up to 30 degrees with RandomRotation.