Convalutional Neural networks

Question 1

What a re convolutional neural networks specifically designed for?

Answer 1

Images

Question 2

What are 4 applications of convolutional neural networks?

Answer 2

> Image analysis
Facial recognition
Autonomous driving
Video analysis

Question 3

What is the main feature that differentiates a convolutional neural network from a normal neural network?

Answer 3

The location of each pixel is taken into account. A normal neural network doesnt care about the position of pixels

Question 4

How does a convolutional neural netowork work?

Answer 4

We apply a filter to the input function and combine the two. We then calculate the result and integrate it

Question 5

What is the standard motion of image filtering?

Answer 5

Applying the image filter over the image and calculating the result.
Shifting a pixel to the right and repeating.
At the end of the row, shifting down a pixel and doing it again.

Question 6

After filtering, now what?

Answer 6

we take the result of the filter and apply it to the neural network

Question 7

What is padding?

Answer 7

The result is often smaller than the image becaus ethe filter restricts the size of the result. Padding adds, data to the outside of the image (either 0s/ 1s or repeating the outer most values). This result is equal to the shape of the original image.

Question 8

What is strides?

Answer 8

This is when you change the motion of filtering.

You can iterate through the image different amounts.

Question 9

What can we do to improve image recognition?

Answer 9

We can use multiple filters

Question 10

What are the different layer types used in image recognition?

Answer 10

> CONV: