augmentations

Question 1

What is the purpose of Rotation in image augmentation?

Answer 1

Rotation changes the orientation of an image to make the model invariant to different viewpoints.

Question 2

How does Translation improve model robustness?

Answer 2

Translation shifts the image along the x or y axis to simulate input in different positions, improving spatial invariance.

Question 3

What is Scaling in image augmentation?

Answer 3

Scaling resizes the image to simulate objects appearing at different distances from the camera.

Question 4

What is the effect of Flipping an image during augmentation?

Answer 4

Flipping horizontally or vertically creates variations in perspective, improving model performance on symmetric data.

Question 5

What does Cropping achieve in image augmentation?

Answer 5

Cropping focuses on specific regions of an image, removing unwanted areas.

Question 6

What is Brightness Adjustment in color transformations?

Answer 6

Brightness Adjustment alters the intensity of light in an image, simulating different lighting conditions.

Question 7

How does Contrast Adjustment help in image augmentation?

Answer 7

Contrast Adjustment changes the difference between light and dark areas, enhancing or reducing detail visibility.

Question 8

What is Saturation Adjustment in augmentation?

Answer 8

Saturation Adjustment modifies the intensity of colors, making them more vibrant or muted.

Question 9

What is the purpose of Hue Adjustment in color transformations?

Answer 9

Hue Adjustment changes the overall color tone of an image to simulate different lighting or environmental conditions.

Question 10

What does Grayscale Conversion do in image augmentation?

Answer 10

Grayscale Conversion removes color information, reducing the image to shades of gray and helping models focus on texture and structure.

Question 11

What is Gaussian Noise Injection?

Answer 11

Gaussian Noise adds random variations to pixel values, simulating sensor noise and improving model robustness.

Question 12

What is Salt-and-Pepper Noise in image augmentation?

Answer 12

Salt-and-Pepper Noise randomly replaces some pixel values with white or black, mimicking transmission errors.

Question 13

What is Speckle Noise and its use in augmentation?

Answer 13

Speckle Noise adds multiplicative noise, often used to simulate distortions in radar or ultrasound imaging.

Question 14

What is Gaussian Blur in image augmentation?

Answer 14

Gaussian Blur smoothens the image by reducing high-frequency content, mimicking out-of-focus or motion blur.

Question 15

What is Motion Blur, and why is it used?

Answer 15

Motion Blur simulates the effect of motion during image capture, enhancing model robustness to such distortions.

Question 16

What is Edge Enhancement in augmentation?

Answer 16

Edge Enhancement sharpens edges in the image, making features like boundaries more pronounced.

Question 17

What is the purpose of Random Erasing in cutout and masking?

Answer 17

Random Erasing removes random regions of the image to improve robustness to missing or occluded features.

Question 18

What is CutMix in augmentation?

Answer 18

CutMix combines patches from multiple images, blending their labels, which encourages the model to learn from combined contexts.

Question 19

What is Hide-and-Seek Augmentation?

Answer 19

Hide-and-Seek hides random parts of the image (more than cutout which is only 1 crop) to force the model to focus on remaining visible areas.

Question 20

What is MixUp in mixing techniques?

Answer 20

MixUp creates linear combinations of two images and their labels, enhancing generalization by introducing interpolated samples.

Question 21

What is AugMix in augmentation?

Answer 21

AugMix blends multiple augmentations into a single image, aiming to increase robustness and uncertainty estimation.