Built-in datasets (torchvision.datasets) in PyTorch - Model Metrics & Evaluation

Metrics & Evaluation - Built-in datasets (torchvision.datasets)

Which metric matters for Built-in datasets (torchvision.datasets) and WHY

When using built-in datasets like those in torchvision.datasets, the key metrics depend on the task you train your model for. For example, if you train a classifier on CIFAR-10, accuracy is a simple and clear metric to see how well your model predicts the correct class.

However, if the dataset is imbalanced (some classes appear more than others), accuracy alone can be misleading. In that case, metrics like precision, recall, and F1 score become important to understand how well the model performs on each class.

In summary, the dataset provides the data, but the metric you choose depends on your model's goal and the dataset's balance.

Confusion matrix example for a 3-class classification

       Predicted
       C1  C2  C3
    T1  50  2   3
    T2  4   45  1
    T3  5   2   48
    

Here, T1, T2, T3 are true classes, and C1, C2, C3 are predicted classes.

This matrix helps calculate precision and recall for each class, showing where the model confuses classes.

Precision vs Recall tradeoff with built-in datasets

Imagine you use the MNIST dataset to train a digit recognizer. If your model is very cautious and only predicts a digit when very sure, it may have high precision (few wrong guesses) but low recall (misses many digits).

On the other hand, if it guesses more often, recall improves (finds more digits), but precision drops (more wrong guesses).

Choosing the right balance depends on your goal. For example, in medical image datasets, missing a disease (low recall) is worse than a false alarm (low precision).

Good vs Bad metric values for models trained on built-in datasets

Good metrics:

• High accuracy (e.g., >90% on CIFAR-10) means the model predicts most images correctly.
• Balanced precision and recall across classes show the model is fair and reliable.
• F1 score close to accuracy indicates no big tradeoff issues.

Bad metrics:

• High accuracy but very low recall on some classes means the model ignores those classes.
• Very low precision means many wrong predictions, confusing the model.
• Large difference between training and validation accuracy suggests overfitting.

Common pitfalls when evaluating models on built-in datasets

• Accuracy paradox: High accuracy can be misleading if classes are imbalanced.
• Data leakage: Accidentally using test data during training inflates metrics falsely.
• Overfitting: Model performs well on training data but poorly on new data.
• Ignoring class imbalance: Not checking per-class metrics hides poor performance on rare classes.

Self-check question

Your model trained on a built-in dataset has 98% accuracy but only 12% recall on a rare class. Is it good for production? Why or why not?

Answer: No, it is not good. The low recall means the model misses most examples of the rare class, which could be critical depending on the task. High accuracy is misleading here because the rare class is small compared to others. You need to improve recall to catch more of the rare class.

Key Result

Accuracy is a simple starting point, but precision, recall, and F1 score reveal true model performance on built-in datasets.

Practice

1. What is the main purpose of using torchvision.datasets in PyTorch?

easy

A. To visualize model training progress

B. To create custom neural network layers

C. To easily download and load popular image datasets

D. To optimize model hyperparameters automatically

Solution

1. Step 1: Understand the role of torchvision.datasets

   It provides built-in access to popular image datasets like CIFAR10, MNIST, etc., handling download and loading.

2. Step 2: Compare with other options

   Options A, B, and D relate to different PyTorch functionalities, not dataset loading.

3. Final Answer:

   To easily download and load popular image datasets -> Option C

4. Quick Check:

   torchvision.datasets = dataset loading [OK]

Hint: Remember: torchvision.datasets = easy dataset access [OK]

Common Mistakes:

• Confusing datasets with model layers
• Thinking it handles training visualization
• Assuming it tunes hyperparameters

2. Which of the following is the correct way to import the CIFAR10 dataset from torchvision?

easy

A. from torchvision.datasets import CIFAR10

B. import torchvision.datasets.CIFAR10

C. from torchvision import CIFAR10

D. import CIFAR10 from torchvision.datasets

Solution

1. Step 1: Recall Python import syntax for modules and classes

   The correct syntax to import a class from a module is: from module import Class.

2. Step 2: Apply to torchvision.datasets and CIFAR10

   So, from torchvision.datasets import CIFAR10 is correct. Other options have syntax errors or wrong order.

3. Final Answer:

   from torchvision.datasets import CIFAR10 -> Option A

4. Quick Check:

   Correct import syntax = from torchvision.datasets import CIFAR10 [OK]

Hint: Use 'from module import Class' for dataset imports [OK]

Common Mistakes:

• Using dot notation in import statement incorrectly
• Swapping import order
• Trying to import dataset directly without module

3. What will be the output shape of images loaded from the MNIST dataset using torchvision.datasets with default settings?

medium

A. (28, 28)

B. (1, 28, 28)

C. (3, 28, 28)

D. (32, 32, 3)

Solution

1. Step 1: Recall MNIST image format

   MNIST images are grayscale with size 28x28 pixels.

2. Step 2: Understand PyTorch tensor shape for images

   PyTorch uses channel-first format: (channels, height, width). Grayscale images have 1 channel, so shape is (1, 28, 28).

3. Final Answer:

   (1, 28, 28) -> Option B

4. Quick Check:

   MNIST images = grayscale 28x28 = (1,28,28) [OK]

Hint: PyTorch images: channels first, grayscale = 1 channel [OK]

Common Mistakes:

• Assuming shape is height x width only
• Confusing RGB (3 channels) with grayscale
• Using (28, 28) without channel dimension

4. Consider this code snippet:

from torchvision.datasets import CIFAR10
train_data = CIFAR10(root='./data', train=True, download=True)
print(len(train_data))

What error will occur if download=True is omitted and data is not already downloaded?

medium

A. SyntaxError: Missing download argument

B. No error, data loads automatically

C. FileNotFoundError: No such file or directory

D. RuntimeError: Dataset not found and download=False

Solution

1. Step 1: Understand CIFAR10 dataset download behavior

   If download=False (default) and data is missing, it raises an error because it won't download automatically.

2. Step 2: Identify the error type

   The error raised is a RuntimeError indicating dataset not found and download flag is False.

3. Final Answer:

   RuntimeError: Dataset not found and download=False -> Option D

4. Quick Check:

   Missing data + download=False = RuntimeError [OK]

Hint: Always set download=True first time to avoid RuntimeError [OK]

Common Mistakes:

• Expecting automatic download without flag
• Confusing RuntimeError with SyntaxError
• Assuming FileNotFoundError is raised

5. You want to load the CIFAR10 training dataset with images normalized to mean=0.5 and std=0.5 for each RGB channel. Which code snippet correctly applies this transform using torchvision?

hard

A. from torchvision.datasets import CIFAR10
from torchvision.transforms import Compose, ToTensor, Normalize
transform = Compose([ToTensor(), Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))])
data = CIFAR10(root='./data', train=True, download=True, transform=transform)

B. from torchvision.datasets import CIFAR10
transform = Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))
data = CIFAR10(root='./data', train=True, download=True, transform=transform)

C. from torchvision.datasets import CIFAR10
from torchvision.transforms import ToTensor
transform = ToTensor()
data = CIFAR10(root='./data', train=True, download=True, transform=transform)
data = Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))

D. from torchvision.datasets import CIFAR10
from torchvision.transforms import Compose, Normalize
transform = Compose([Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))])
data = CIFAR10(root='./data', train=True, download=True, transform=transform)

Solution

1. Step 1: Understand required transforms for normalization

   Images must first be converted to tensors with ToTensor() before normalization.

2. Step 2: Check transform composition

   Use Compose to chain ToTensor() and Normalize() in order.

3. Step 3: Verify each option

   from torchvision.datasets import CIFAR10 from torchvision.transforms import Compose, ToTensor, Normalize transform = Compose([ToTensor(), Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))]) data = CIFAR10(root='./data', train=True, download=True, transform=transform) correctly composes both transforms and passes to dataset. Others miss ToTensor() or apply transforms incorrectly.

4. Final Answer:

   Option A code snippet with Compose, ToTensor, Normalize -> Option A

5. Quick Check:

   Normalize after ToTensor using Compose = from torchvision.datasets import CIFAR10 from torchvision.transforms import Compose, ToTensor, Normalize transform = Compose([ToTensor(), Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))]) data = CIFAR10(root='./data', train=True, download=True, transform=transform) [OK]

Hint: Always use ToTensor before Normalize in Compose [OK]

Common Mistakes:

• Applying Normalize without ToTensor first
• Not using Compose to chain transforms
• Applying transforms outside dataset argument