Eng 133 check 2

How are Sobel kernels different from Gaussian kernels?

Sobel kernels are used for edge detection, while gaussian kernels are used to smooth and blur

What is the x-gradient image?

It shows horizontal changes in intensity, and highlights vertical edges.

What is the y-gradient image?

It shows vertical changes in intensity, and highlights horizontal edges.

What is meant by gradient?

Gradient is a measure of the intensity change at each pixel.

What is the Sobel filter detecting?

It detects edges through pixel brightness

Why do we apply a threshold?

to create a cutoff line to find the real edges of an image

What happens if the threshold is too high? What happens if the threshold is too low?

Too high - edges are missed and lost

Too low - too much background noise is kept

Why do we want to detect the edges of the image?

Reduce the size of image, but still keeps all important features

What features are extracted from the color image?

the specific RGB / HSV channel values

What features are extracted from the grayscale image?

gradients, edges, texture

What filters do you use to detect circles?

Hough Circle Transform

What filters do you use to detect triangles?

cv2.findContours and cv2.approxPolyDP

What filters do you use to count lines?

Hough Line Transform

What is the metadata CSV file? What data does it contain?

It maps each image to its identity using

‘Path’ and ‘ClassID’

What is the purpose of extracting features from our images?

to turn the pixels in numericals that programs can digest easily

What is the features CSV file? What data does it contain?

It stores all calculations from the images likes (means, stds, has_circle and has_triangle

What is feature scaling? Why is it useful?

It rescales values into a range so there are no disproportionately large ranges taking over

What is the feature space? Why is it useful?

A space where each dimension correlates to a different feature. It useful to separate data classes

how do the feature plots change when you apply feature scaling?

The shape stays the same, but the axes changes to make all values inside a certain range

Why do the plots with has_triangles and has_circles look so different to the plots comparing the color features?

Shape features use binary and create rigid rows, while color features are spread across a smooth spectrum