CSE 163

Convolutions�

Suh Young Choi�

🎶 Listening to: Inception soundtrack

💬 Before Class: What’s been your favorite before-class music so far?

Announcements

Project Proposal open now on Gradescope

• Make sure you have access!
• Group-finding form can be found in Ed post #249 (this closes tonight)
• Full spec can be found on the course website

​

HW5 and LR5 due on Thursday (last one of the quarter!)

​

2

This Time

• Convolutions
• Kernel operations

Last Time

• numpy arrays
• Broadcasting
• Images

3

Images as

Matrices

​

​

4

Grey-scale images can be represented as matrices.

Grey-scale: 255

Grey-scale: 0

data = imageio.imread(‘...’)�

data[rows, columns] = #

Color Images

When you overlap each color channel, it creates a picture we are used to seeing.

• Pixels on your monitor let out specified R/G/B light

5

data[rows, columns, channels] = #

Convolution

When wanting to use “local” information, we commonly use a sliding window approach (i.e. a convolution)

​

Move the sliding window across the image, and compute the sum of the element wise product of the window (kernel) and image

6

Image

Kernel

Convolution Example

7

Common Kernels

What do the numbers in the kernel do?

8

Identity

Edge Detection

Sharpen

Box Blur

Image Classification

For a really long time, image classification was done by painstakingly crafting these features (like edge detectors), by hand.

​

This kind of worked, but we quickly hit our peak using this method.

​

Then came the buzz-word… deep learning

9

Image Classification

• Is this a solved problem?
• We get pretty decent error rates on challenges like ImageNet
• What we can’t do
• Sometimes can’t generalize to other real-world datasets
• Adversarial attacks

10

LR topics may include:

• How convolutions work
• Kernel operations

11

Before Next Time

• Complete Lesson 20
• Remember not for points, but do go towards Checkpoint Tokens
• HW5 and LR5 due tomorrow
• Go to section!

Next Time

• Machine learning with images

​

12