L01 Slides

Pattern Recognition
A. G. Schwing
University of Illinois at Urbana-Champaign, 2017
A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 1 / 37

L1: ECE 544 Pattern Recognition

Goals of this lecture

Introduction to pattern recognition

Organizational details

Style of this lecture

Style of this lecture

Interactive

Pattern recognition

somewhat dated joke

What is pattern recognition? What applications?
Discuss with your neighbor

Where pattern recognition shows up:

Chess play (Deep Blue, IBM, 1997)

Jeopardy (Watson, IBM, 2011)

Playing Atari games (DQN, Deepmind, 2015)

Game of Go (AlphaGo, Deepmind, 2016)

Game of Go (AlphaGo, Deepmind, 2016)
Autonomous Driving (Audi, BMW, Mercedes, Uber, Waymo, Lyft,
NVIDIA, Intel, ...)

Example:
Let’s get a computer to recognize whether there is a cat in the image.

Example:
Algorithm/Model

Example:
Algorithm/Model yes

Example:
Algorithm/Model

Example:
Algorithm/Model yes

Example:
Algorithm/Model

Example:
Algorithm/Model no

Formulation of pattern recognition:

input value: x (i)

input value: x (i)
label/output: y (i)

input value: x (i)
label/output: y (i)
x (i) Model/Algorithm y (i)

input value: x (i)
label/output: y (i)
Inference

input value: x (i)
label/output: y (i)
Inference
Prediction


Model/Algorithm depends on parameters w


Learning/Fitting of parameters w


Learning/Fitting of parameters w
Based on dataset D = {(x (i) , y (i) )}N
i=1

How could we recognize a cat? What should the algorithm do?

How does an image look like for a computer?

Demo

Demo

Cell phone screen:

Given that we know how a computer ‘sees’ an image, how can it
recognize cats?

recognize cats?

recognize cats?
Still very hard to describe what a cat looks like.

recognize cats?
Still very hard to describe what a cat looks like.
Let’s look at tons of examples (Dataset).

Dataset (Thousands of examples):

Instead of asking to recognize cats in general, let’s recognize cats in

this dataset


this dataset


this dataset
Algorithm: if bottom left corner is black (0) say no, otherwise say yes


Algorithm/Model

Algorithm/Model yes

Algorithm/Model

Algorithm/Model no

Algorithm/Model no
Works perfect on our dataset. :)

Algorithm/Model

Algorithm/Model yes (?)

Algorithm/Model yes (?)

But does not generalize to other data. :(

Conclusion:

Conclusion:
We designed a simple “classifier” that works on this dataset but

doesn’t work on real data.

Conclusion:

To our rescue:

Conclusion:

To our rescue:
Machine learning/Pattern recognition found mechanisms to search for

mappings which generalize.

Conclusion:

To our rescue:
Machine learning/Pattern recognition found mechanisms to search for

mappings which generalize.
Scope of this class:
In this lecture we talk about mechanisms. A detailed treatment about

generalization is left to lectures on learning theory.

Categorization of pattern recognition algorithms according to

Available annotated data (supervised vs. unsupervised)

Complexity of model (linear vs. non-linear)

Structure of output (independent vs. structured)

Structure of output (independent vs. structured)
Modeling of data (x (i) ) or label (y (i) ) (generative vs. discriminative)

Syllabus

Classification tasks:

Spam filter

Spam filter
Cancer classification

Spam filter
Cancer classification
...

More examples

Example: Differentiating between 1000 image categories

Which object is illustrated?


Car


Car
Truck


Car
Truck
Recreational Vehicle


Car
Truck
Ambulance truck


Car
Truck
Ambulance truck
Fire truck

Example: Instance level video segmentation

Results from A. Karpathy & L. Fei-Fei
Example: Describing an image
Works

Works

Works, ...

Works, ...

Kind of works, ...

Kind of works, ...

And fails occasionally too

And fails occasionally too

Creativity


Object Labels: ‘person’ Object Labels: ‘person’,‘bus’
AG-CVAE sentences: AG-CVAE sentences:
a man and a woman standing in a room a man and a woman sitting on a bus
a man and a woman are playing a game a man and a woman sitting on a train
a man standing next to a woman in a room a man and woman sitting on a bus
a man standing next to a woman in a field a man and a woman sitting on a bench
a man standing next to a woman in a suit a man and a woman are sitting on a bus
Object Labels: ‘person’, ‘remote’ Object Labels: ‘person’, ‘train’
AG-CVAE sentences: AG-CVAE sentences:
a man and a woman playing a video game a man and a woman sitting on a train
a man and a woman are playing a video game a woman and a woman sitting on a train
a man and woman are playing a video game a woman sitting on a train next to a train
a man and a woman playing a game with a remote a woman sitting on a bench in a train
a woman holding a nintendo wii game controller a man and a woman sitting on a bench

Results from U. Jain et al.
Example: Asking questions about an image

Results from U. Jain et al.
Example: Asking questions about an image

Questionnaire
Syllabus
Assignments
Project
Final exam
Piazza
Grading policy (50% Final, 30% Homework, 20% Project)

So far:
What can a computer do for you?

So far:
What can a computer do for you?
Now:
What will a computer be able to do for you?

Future: (years from 2016; from https://arxiv.org/pdf/1705.08807.pdf)

Videos

Amazing future ahead. Let’s make it happen.

L01 Slides

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

L01 Slides

Uploaded by

Copyright:

Available Formats

Pattern Recognition

University of Illinois at Urbana-Champaign, 2017

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 1 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 2 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 3 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 3 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 3 / 37

Style of this lecture

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 3 / 37

Style of this lecture

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 3 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 4 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 5 / 37

Discuss with your neighbor

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 6 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 7 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 7 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 7 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 7 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 7 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 7 / 37

Let’s get a computer to recognize whether there is a cat in the image.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 8 / 37

Let’s get a computer to recognize whether there is a cat in the image.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 8 / 37

Let’s get a computer to recognize whether there is a cat in the image.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 8 / 37

Let’s get a computer to recognize whether there is a cat in the image.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 8 / 37

Let’s get a computer to recognize whether there is a cat in the image.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 8 / 37

Let’s get a computer to recognize whether there is a cat in the image.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 8 / 37

Let’s get a computer to recognize whether there is a cat in the image.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 8 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 9 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 9 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 9 / 37

x (i) Model/Algorithm y (i)

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 9 / 37

x (i) Model/Algorithm y (i)

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 9 / 37

x (i) Model/Algorithm y (i)

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 9 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 10 / 37

Model/Algorithm depends on parameters w

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 10 / 37

Model/Algorithm depends on parameters w

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 10 / 37

Model/Algorithm depends on parameters w

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 10 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 11 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 12 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 12 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 12 / 37

Cell phone screen:

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 13 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 14 / 37

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 14 / 37

Still very hard to describe what a cat looks like.

A. G. Schwing (UofI) ECE 544: Pattern Recognitition 2017 14 / 37

Still very hard to describe what a cat looks like.

Let’s look at tons of examples (Dataset).