Top-K accuracy in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the function that calculates top-k accuracy from PyTorch.

Computer Vision

from torchmetrics import [1]

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

AAccuracy

BRecall

CPrecision

DTopKAccuracy

Attempts:

3 left

💡 Hint

Common Mistakes

Choosing TopKAccuracy (does not exist)

Using Precision or Recall which are different metrics

The Accuracy class from torchmetrics calculates the top-k accuracy metric.

2fill in blank

medium

Complete the code to create an Accuracy metric object for top 3 accuracy.

Computer Vision

top3_acc = Accuracy(task="multiclass", top_k=[1])

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

A1

B3

C5

D10

Attempts:

3 left

💡 Hint

Common Mistakes

Using k=1 which is normal accuracy

Using k=5 or 10 which are valid but not for top 3 accuracy

Setting top_k=3 means the metric checks if the true label is in the top 3 predictions.

3fill in blank

hard

Fix the error in the code to compute top-5 accuracy from model outputs and labels.

Computer Vision

top5_acc = Accuracy(task="multiclass", top_k=[1])
outputs = torch.randn(8, 1000)  # batch of 8, 1000 classes
labels = torch.randint(0, 1000, (8,))
accuracy = top5_acc(outputs, labels)

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

A10

B3

C1

D5

Attempts:

3 left

💡 Hint

Common Mistakes

Using k=1 or k=3 which do not compute top-5 accuracy

Passing wrong shapes for outputs or labels

Top-5 accuracy requires top_k=5 to check if true label is in top 5 predictions.

4fill in blank

hard

Fill both blanks to create a dictionary of top-1 and top-3 accuracy metrics.

Computer Vision

metrics = {
    'top1': Accuracy(task="multiclass", top_k=[1]),
    'top3': Accuracy(task="multiclass", top_k=[2])
}

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

A1

B2

C3

D5

Attempts:

3 left

💡 Hint

Common Mistakes

Confusing top-1 with top-3 values

Using k=2 which is not standard for top-k accuracy

Top-1 accuracy uses top_k=1 and top-3 accuracy uses top_k=3.

5fill in blank

hard

Fill all three blanks to compute and print top-1, top-3, and top-5 accuracy from outputs and labels.

Computer Vision

outputs = torch.randn(16, 1000)
labels = torch.randint(0, 1000, (16,))
metrics = {
    'top1': Accuracy(task="multiclass", top_k=[1]),
    'top3': Accuracy(task="multiclass", top_k=[2]),
    'top5': Accuracy(task="multiclass", top_k=[3])
}
for name, metric in metrics.items():
    acc = metric(outputs, labels)
    print(f"{name} accuracy: {acc.item():.4f}")

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

A1

B3

C5

D10

Attempts:

3 left

💡 Hint

Common Mistakes

Mixing up k values for different top-k metrics

Using invalid k values like 10 for top1 or top3

Top-1, top-3, and top-5 accuracies require top_k values 1, 3, and 5 respectively.