Lightning AI Studios: Never set up a local environment again →

Log in or create a free Lightning.ai account to track your progress and access additional course materials  

Deep Learning Fundamentals

  • Pages

    • Final certification exam

    Deep Learning Fundamentals > Unit 7 > Unit 7.5
    Course Progress:

    Unit 7.5 Improving Predictions with Data Augmentation

    Slides

    References

    Code

    • Parts 2 and 3, 7.5-data-aug/

    What we covered in this video lecture

    In this series of lecture and coding videos, we explored various different image augmentation techniques. Image augmentation is a convenient approach that allows us to reduce overfitting and increase the generalization performance of our models by generating a larger variety of our existing training examples.

    Additional resources if you want to learn more

    There are many other image augmentation libraries that extend the functionality of torchvision. For example, I recommend checking out Image Albumentations.

    Log in or create a free Lightning.ai account to access:

    • Quizzes
    • Completion badges
    • Progress tracking
    • Additional downloadable content
    • Additional AI education resources
    • Notifications when new units are released
    • Free cloud computing credits

    Quiz: 7.5 Improving Predictions with Data Augmentation - Part 1

    How does data augmentation help improve the performance of machine learning models?

    Incorrect. Data augmentation does not affect the number of layers in a model.

    Incorrect. Data augmentation does not directly increase a model’s capacity; it helps the model generalize better by providing more diverse training data.

    Correct. And this helps the model generalize better by providing more diverse training data.

    Incorrect. Data augmentation does not affect the number of trainable parameters.

    Please answer all questions to proceed.

    Quiz: 7.5 Improving Predictions with Data Augmentation - Part 2

    Which of the following is NOT a common data augmentation technique for image data?

    Incorrect. We used rotation in the lecture.

    Incorrect. We used horizontal flipping in the lecture.

    Incorrect. We used padding in the lecture.

    Correct. Stemming is a text preprocessing technique used in natural language processing, not image data augmentation. It involves reducing words to their root form.

    Please answer all questions to proceed.

    Quiz: 7.5 Improving Predictions with Data Augmentation - Part 3

    What is a potential downside of data augmentation?

    Incorrect. Data augmentation generally helps reduce overfitting, not cause underfitting.

    Correct. This is because additional preprocessing steps are needed to generate the augmented data and because we potentially have to train the model for more epochs.

    Incorrect. Data augmentation is designed to improve the model’s ability to generalize by providing more diverse training data.

    Incorrect. Data augmentation does not directly affect the interpretability of a model.

    Please answer all questions to proceed.
    Watch Video 1

    Unit 7.5

    Videos
    Follow along in a Lightning Studio

    DL Fundamentals 7: Computer Vision

    Sebastian
    Launch Studio →