Geometric transforms (rotate, flip, crop) in Computer Vision - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is the purpose of geometric transforms like rotate, flip, and crop in computer vision?

They help change the position or shape of images to augment data, improve model robustness, or focus on important parts of the image.

beginner

How does rotating an image affect its pixel positions?

Rotating an image moves each pixel around a center point by a certain angle, changing their coordinates but keeping the image content intact.

beginner

What is the difference between horizontal flip and vertical flip?

Horizontal flip mirrors the image left to right, while vertical flip mirrors it top to bottom.

beginner

Why is cropping useful in image processing?

Cropping cuts out a part of the image to focus on important details or remove unwanted areas, which can help models learn better.

intermediate

What should you be careful about when applying geometric transforms during training?

Ensure transforms do not distort important features or labels, and keep the image size consistent if required by the model.

Which geometric transform flips an image top to bottom?

A Vertical flip

B Rotate 90 degrees

C Horizontal flip

D Crop

Vertical flip mirrors the image along the horizontal axis, flipping it top to bottom.

What happens to pixel positions when you rotate an image by 90 degrees?

A Pixels stay in the same place

B Pixels move around the center point

C Pixels are flipped horizontally

D Pixels are cropped

Rotation moves pixels around the image center by the rotation angle.

Why do we crop images in computer vision tasks?

A To focus on important parts

B To add noise

C To flip the image

D To rotate the image

Cropping helps focus on important details or remove irrelevant parts.

Which transform would you use to create a mirror image left to right?

A Vertical flip

B Rotate 180 degrees

C Horizontal flip

D Crop

Horizontal flip mirrors the image along the vertical axis.

What is a key caution when applying geometric transforms during training?

A Transforms always improve accuracy

B Transforms remove labels

C Transforms reduce image size

D Transforms can distort important features

Transforms can distort features if not applied carefully, affecting model learning.

Explain how rotating, flipping, and cropping images help in training computer vision models.

Think about how changing images can help models see more variety.

Describe the difference between horizontal flip and vertical flip and when you might use each.

Imagine flipping a photo like a page in a book or turning it upside down.