Built-in datasets (torchvision.datasets) in PyTorch - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of loading CIFAR10 dataset

What is the output of the following code snippet when loading the CIFAR10 training dataset?

PyTorch

import torchvision.datasets as datasets
cifar10_train = datasets.CIFAR10(root='./data', train=True, download=False)
print(len(cifar10_train))

A 6000

B 60000

C 5000

D 50000

💡 Hint

Recall the number of training images in CIFAR10 dataset.

The CIFAR10 training dataset contains exactly 50,000 images. The code prints the length of the dataset, which is 50000.

❓ Model Choice

intermediate

Choosing dataset for handwritten digit recognition

Which torchvision built-in dataset is best suited for training a model to recognize handwritten digits?

A datasets.CIFAR10

B datasets.MNIST

C datasets.FashionMNIST

D datasets.ImageNet

💡 Hint

Think about which dataset contains handwritten digits.

MNIST contains grayscale images of handwritten digits 0-9, making it ideal for digit recognition tasks.

❓ Hyperparameter

advanced

Effect of 'transform' parameter in torchvision datasets

What is the effect of passing a torchvision.transforms.Compose object to the 'transform' parameter when loading a dataset like CIFAR10?

A It applies the specified transformations to each image when accessed from the dataset.

B It changes the dataset size by filtering images based on the transform.

C It downloads additional images matching the transform criteria.

D It converts the dataset labels to one-hot encoded vectors.

💡 Hint

Consider when and how transforms are applied in PyTorch datasets.

The 'transform' parameter applies the given transformations to each image on-the-fly when the image is accessed, without changing dataset size or labels.

❓ Metrics

advanced

Number of classes in FashionMNIST dataset

How many distinct classes does the FashionMNIST dataset contain?

A 50

B 100

C 10

D 5

💡 Hint

FashionMNIST has the same number of classes as MNIST.

FashionMNIST contains 10 classes representing different clothing items.

🔧 Debug

expert

Error raised when loading ImageNet without download

What error will be raised by the following code if the ImageNet dataset is not downloaded in the specified root folder? import torchvision.datasets as datasets imagenet = datasets.ImageNet(root='./data', split='train', download=False)

A RuntimeError

B FileNotFoundError

C ValueError

D TypeError

💡 Hint

Think about what happens if the dataset files are missing and download is False.

If the ImageNet dataset files are not found in the root directory and download is False, a RuntimeError is raised.