Custom detection dataset in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the PyTorch Dataset class.

PyTorch

from torch.utils.data import [1]

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

ADataLoader

BTensor

CDataset

DModule

Attempts:

3 left

💡 Hint

Common Mistakes

Importing DataLoader instead of Dataset

Importing Tensor or Module which are unrelated here

We import Dataset to create a custom dataset class in PyTorch.

2fill in blank

medium

Complete the code to initialize the dataset with image paths and annotations.

PyTorch

class CustomDataset(Dataset):
    def __init__(self, [1], annotations):
        self.images = images
        self.annotations = annotations

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

Adata

Bself

Clabels

Dimages

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'self' as a parameter name

Using 'labels' or 'data' which are not the parameter names here

The constructor takes images and annotations as inputs.

3fill in blank

hard

Fix the error in the __len__ method to return dataset size.

PyTorch

def __len__(self):
    return [1](self.images)

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

Alen

Bsize

Ccount

Dlength

Attempts:

3 left

💡 Hint

Common Mistakes

Using count() which is not a built-in function for lists

Using size() or length() which are not valid Python functions

The len() function returns the number of items in a list.

4fill in blank

hard

Fill both blanks to load an image and its annotation in __getitem__.

PyTorch

def __getitem__(self, idx):
    image = Image.open(self.images[[1]])
    annotation = self.annotations[[2]]
    return image, annotation

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

Aidx

B0

C1

Dself

Attempts:

3 left

💡 Hint

Common Mistakes

Using 0 or 1 instead of idx, which always returns the first or second item

Using 'self' which is invalid as an index

We use the index idx to get the image path and annotation at the same position.

5fill in blank

hard

Fill all three blanks to apply a transform to the image if provided.

PyTorch

def __getitem__(self, idx):
    image = Image.open(self.images[idx])
    annotation = self.annotations[idx]
    if [1] is not None:
        image = [2](image)
    return [3], annotation

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

Aself.transform

Btransform

Cimage

Dself

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'transform' without self, which is undefined

Returning the original image variable name incorrectly

We check if self.transform exists, then apply it to the image, and return the transformed image.