Learn Before
Write an essay explaining how simulated motion blur helps align cat detector datasets
Question: Explain the process and rationale of adding simulated motion blur to non-blurry internet training images in a cat detector system. In your response, explain the root cause of the difference between the training and dev sets and how this synthesis addresses it based on the ML Yearning principles.
Sample answer: In the cat detector project, the dev set contains motion-blurry images because they are taken by cellphone users who slightly move their phones. Conversely, the training set contains clear internet images. To align the two sets, developers can synthesize data by adding simulated motion blur to the clear internet images. This reduces the distribution mismatch and helps the model generalize better to the dev set.
Key points:
- Dev set images have motion blur from cellphone users moving their phones.
- Training images are clear internet images.
- Simulated motion blur is added to training images to make them more similar to the dev set.
Rubric: Answers must explain that the dev set has motion blur due to cellphone users moving their phone, the training set comes from clear internet images, and simulated motion blur is added to make training images resemble the dev set to close the distribution gap.
0
1
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Related
In the cat image detector example, why do dev set images tend to have more motion blur than training images?
True or False: Adding simulated motion blur to non-blurry training images can reduce the distribution mismatch between training and dev sets.
To make training images more similar to a dev set with motion blur, you can add simulated _____ to non-blurry training images.
Why is simulated motion blur added to non-blurry training images in the cat detector example?
Motion blur in dev-set cat images is caused by cellphone users slightly moving their phone while taking pictures.
To close the distribution gap in the cat detector, you add _____ to non-blurry training images.
Match each dataset or image source to its characteristic in the cat detector example.
Order the steps to apply simulated motion blur to improve a cat detector's performance on a blurry dev set.
What specific distribution gap does adding simulated motion blur to training images address in the cat detector example?
Internet images used as cat detector training data typically have the same level of motion blur as cellphone-captured dev-set images.
In the cat detector example, the non-blurry training images that receive simulated blur originally come from _____ images.
Match each observation in the cat detector scenario to the corresponding explanation or action.
Order the reasoning steps that lead to the decision to use simulated motion blur as an artificial data synthesis technique.
Write an essay explaining how simulated motion blur helps align cat detector datasets
Diagnose and resolve the dataset distribution gap in a mobile cat detector
Explain why simulated motion blur is added to clear training images