CNN CNN - Convolutional Neural Networks Data Augmentation Techniques Questions and Answers 2

Question 1

What does the ColorJitter augmentation randomly alter in a training image?

Accepted Answer

Brightness, contrast, saturation, and hue

Answer

Image geometry

Answer

Image resolution

Answer

Aspect ratio

Question 2

What is RandAugment?

Accepted Answer

An automated augmentation policy that uniformly samples from a set of augmentation operations

Answer

A random weight initialization scheme

Answer

A random architecture search method

Answer

A random dropout strategy

Question 3

What is Test-Time Augmentation (TTA) used for in CNN inference?

Accepted Answer

Applying multiple augmentations to test images and averaging predictions to improve accuracy

Answer

Speeding up inference time

Answer

Augmenting the test dataset permanently

Answer

Reducing model parameters at test time

Question 4

What does AutoAugment do in the context of CNN training?

Accepted Answer

Searches for an optimal augmentation policy using reinforcement learning

Answer

Automatically selects the best optimizer

Answer

Automatically adjusts learning rate

Answer

Generates new architectures automatically

Question 5

What is the 'Random Erasing' augmentation technique?

Accepted Answer

Randomly selecting a rectangle in the image and replacing it with random pixel values

Answer

Erasing random layers from the network

Answer

Removing random images from the dataset

Answer

Erasing random weight connections

Question 6

What risk does overly aggressive data augmentation introduce during CNN training?

Accepted Answer

It can create unrealistic training samples that hurt model performance

Answer

It always improves accuracy

Answer

It speeds up convergence too much

Answer

It reduces memory usage