MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory...
Transcript of MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory...
![Page 1: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/1.jpg)
MC949 - Computer Vision Lecture #7
Prof. Dr. Anderson Rocha
Microsoft Research Faculty Fellow Affiliate Member, Brazilian Academy of Sciences
Reasoning for Complex Data (Recod) Lab. Institute of Computing, University of Campinas (Unicamp)
Campinas, SP, Brazil
[email protected] http://www.ic.unicamp.br/~rocha
![Page 2: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/2.jpg)
Machine Learning: Introductory Lesson
This lecture slides were made based on slides of several researchers such as James Hayes, Derek Hoiem, Alexei Efros, Steve Seitz, David Forsyth and many others. Many thanks to all of these authors.
Reading: PRML, Chapters 1, 6 and 7
![Page 3: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/3.jpg)
The blue and green colors are actually the same
![Page 4: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/4.jpg)
h3p://blogs.discovermagazine.com/badastronomy/2009/06/24/the-‐blue-‐and-‐the-‐green/
![Page 5: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/5.jpg)
Machine Learning: Overview
Slides: Isabelle Guyon, Erik Sudderth, Mark Johnson, Derek Hoiem
Photo: CMU Machine Learning - Department protests G20
![Page 6: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/6.jpg)
Machine learning: Overview
• Core of ML: Making predicHons or decisions from Data.
• This overview will not go in to depth about the staHsHcal underpinnings of learning methods. We’re looking at ML as a tool. Take MC906 (AI) or MC886 (ML).
![Page 7: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/7.jpg)
Impact of Machine Learning
• Machine Learning is arguably the greatest export from compuHng to other scienHfic fields.
![Page 8: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/8.jpg)
Machine Learning ApplicaHons
Slide: Isabelle Guyon
![Page 9: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/9.jpg)
Image CategorizaHon
Training Labels
Training Images
Classifier Training
Training
Image Features
Trained Classifier
Slide: Derek Hoiem
![Page 10: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/10.jpg)
Image CategorizaHon
Training Labels
Training Images
Classifier Training
Training
Image Features
Image Features
Testing
Test Image
Trained Classifier
Trained Classifier Outdoor
Prediction
Slide: Derek Hoiem
![Page 11: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/11.jpg)
Claim: The decision to use machine learning is more important than the choice of a par'cular learning method.
If you hear somebody talking of a specific learning mechanism, be wary (e.g. YouTube comment "Oooh, we could plug this in to a Neural network and blah blah blah“)
![Page 12: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/12.jpg)
Example: Boundary DetecHon
• Is this a boundary?
![Page 13: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/13.jpg)
Image features
Training Labels
Training Images
Classifier Training
Training
Image Features
Trained Classifier
Slide: Derek Hoiem
![Page 14: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/14.jpg)
General Principles of RepresentaHon • Coverage
– Ensure that all relevant info is captured
• Concision – Minimize number of features without sacrificing coverage
• Directness – Ideal features are independently useful for predicHon
Image Intensity
Slide: Derek Hoiem
![Page 15: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/15.jpg)
Image representaHons
• Templates – Intensity, gradients, etc.
• Histograms – Color, texture, SIFT descriptors, etc.
Slide: Derek Hoiem
![Page 16: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/16.jpg)
Classifiers
Training Labels
Training Images
Classifier Training
Training
Image Features
Trained Classifier
Slide: Derek Hoiem
![Page 17: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/17.jpg)
Learning a classifier Given some set of features with corresponding labels, learn a funcHon to predict the labels from the features
x x
x x
x
x x
x
o o
o
o
o
x2
x1 Slide: Derek Hoiem
![Page 18: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/18.jpg)
One way to think about it…
• Training labels dictate that two examples are the same or different, in some sense
• Features and distance measures define visual similarity
• Classifiers try to learn weights or parameters for features and distance measures so that visual similarity predicts label similarity
Slide: Derek Hoiem
![Page 19: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/19.jpg)
Slide: Erik Sudderth
![Page 20: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/20.jpg)
Dimensionality ReducHon
• PCA, ICA, LLE, Isomap
• PCA is the most important technique to know. It takes advantage of correlaHons in data dimensions to produce the best possible lower dimensional representaHon, according to reconstrucHon error.
• PCA should be used for dimensionality reducHon, not for discovering pa3erns or making predicHons. Don't try to assign semanHc meaning to the bases.
![Page 21: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/21.jpg)
Many classifiers to choose from • SVM • Neural networks • Naïve Bayes • Bayesian network • LogisHc regression • Randomized Forests • Boosted Decision Trees • K-‐nearest neighbor • RBMs • Etc.
Which is the best one?
Slide: Derek Hoiem
![Page 22: MC949 - Computer Visionrocha/teaching/2012s1/mc949/aulas/20… · Machine Learning: Introductory Lesson This lecture slides were made based on slides of several researchers such as](https://reader034.fdocuments.in/reader034/viewer/2022052005/6018d34f855f3c781a6f0235/html5/thumbnails/22.jpg)
Next Two Lectures:
• Monday we'll talk about clustering methods (k-‐means, mean shif) and their common usage in computer vision -‐-‐ building "bag of words” representaHons inspired by the NLP community. We'll be using these models for the next project.
• Wednesday we'll focus specifically on classificaHon methods, e.g. nearest neighbor, naïve-‐Bayes, decision trees, linear SVM, Kernel methods. We’ll use these for the next projects.