C4W1

Question 1

Types of convolutions

Answer 1

Vertical (detect vertical edges on the image)
Horizontal (detect horizontal edges on the image)

Question 2

Alternative to the hand coded convolution filter/kernel

Answer 2

Let the model learn the filter

Question 3

Advantage of letting model to learn the convolutional filter

Answer 3

Model can learn complex filter to detect 45• edges, for example

Question 4

Padding in convolutional NN

Answer 4

Adding additional border of P(padding amount) px around the image

Question 5

Which problem padding solve?

Answer 5

Increase impact of the edge image pixels

Question 6

Valid convolution vs same convolution

Answer 6

Valid - no padding
Same - output size is same as input size

Question 7

What is strided convolution?

Answer 7

When you move convolutional by more than 1 pixel at a time

Question 8

How to do convolution over RGB image?

Answer 8

Have separate filters for each channel

Question 9

How to detect several feature on a image?

Answer 9

Apply filter for detecting each feature and than stack the outputting images so the number of channels in the output will be the same a the number of features you are detecting

Question 10

Where most of the parameters are present?

Answer 10

In a fully connected layer