CVPR2010: Semi-supervised Learning in Vision: Part 1: Introduction

Professor Horst Cerjak, 19.12.20051

Horst Bischof SSL CVPR Tutorial

CVPR 2010 Tutorial Semi-Supervised Learning in

Vision

Inst. for Computer Graphics and VisionGraz University of Technology

A, Saffari, Ch. Leistner, H. Bischof

http://www.icg.tugraz.at/Members/Saffari/ssl-cvpr2010

Typical Vision Tasks/Trends

Internet: Image Search/Classification Categorization

Huge Amounts of data, Huge labeling effort

Internet: Various Video and Image Databases

Huge Amounts of data, Partially/weakly labeled data

Object Categorization

Huge Amounts of data, Huge labeling effort

VisipediaPerona 2009

Surveillance: On-line data/Detection-Tracking

Huge Amounts of data, On-line processing, Scene adaptation

Many other tasks:• Tracking: On-line adaptation to object• Interactive segmentation: Changing model on

the fly• Interactive labeling: Suggestions as you label• …..

Requirements

Huge datasets

Unreliable (partial) labels

Changing environme

Requirements on Learner

On-line learning

RobustnessSemi-

supervised Learning

Theory/Methods Covered

1. Semi-Supervised Learning2. Self-Training3. Generative Models4. Margin Assumption5. Cluster and Manifold Assumption6. Multi-View Learning7. Large-Scale, Multi-Class SSL and Online Learning8. Transfer Learning, Domain Adaptation, and Weakly

Related Data9. Multiple-Instance Learning

Applications Covered

1. Object Detection2. Categorization3. Tracking4. Activity Recognition5. Segmentation

Using:BoostingRandom Forests

Learning Tasks

• Unsupervised learning– Density estimation, Clustering, Dimensionality reduction.

• Semi-Supervised Clustering– Clustering with pair-wise constraints: must-link, cannot-link.

• Semi-Supervised Classification and Regression

• Supervised Learning– Classification, Regression.

How do we get labels?Labels

Semi-Supervised LearningSSL is Supervised Learning...

Goal: Estimate P(y|x) from Labeled DataDl={ (xi,yi) }

But: Additional Source tells about P(x)(e.g., Unlabeled Data Du={xj})

The Interesting Case:

)()()|()|(

xPypyxpxyp =

Supervised learning

Maximum margin

Can Unlabeled Data Help?

low densityaround

decisionboundary

SSL is biologically plausible

• Co-training by infants [Bahrick et.al. 2002]• Human change model once they see unlabeled data

[Zahki et.al. 2007,Zhu et.al. 2007,Vandist et.al. 2007]• Humans do On-line SSL [Zhu ICML 2010]

1. Unlabelled data is cheap/free

2. Labeled data is hard to get

• human annotation is boring• labels may require experts• labels may require special devices• your graduate student is on vacation

Online Learning in Perspective

Memory

0 ∞0

Online

Batch/Off-line

Incremental

Why on-line learning?

Too much training data to fit in memory– Internet!!!

Sample generation process– Tracking, Co-Training

Changing processes– Changing Environment

Why on-line learning?

Specializing– Forget irrelevant information– Specialize to current scene

Interactive Applications– Data labeling– Classifier Training– Specializing (Human in the loop)– Interactive Training (Segmentation)

Robustness in Learning

Noisy input data

Label noise• Semi-supervised learning• Weakly-labeled data• Co-training• Self-learning• On-line learning

Amir SaffariTheory

Christian LeistnerAlgorithms/Applications

CVPR2010: Semi-supervised Learning in Vision: Part 1: Introduction

Education

An Ensemble of Transfer, Semi-supervised and Supervised ...An Ensemble of Transfer, Semi-supervised and Supervised Learning Methods for Pathological Heart Sound Classiﬁcation Ahmed

Semi-supervised Mesh Segmentation and Labeling · Semi-supervised Mesh Segmentation and Labeling Jiajun Lv, ... and potentially more topologi- ... Hujun Bao / Semi-supervised Mesh

Supervised Nonlinear Factorizations Excel In Semi-supervised … · 2020. 12. 9. · Supervised Nonlinear Factorizations Excel In Semi-supervised Regression Josif Grabocka 1, Erind

Inductive Semi-supervised Learning

Learning Semi-Supervised Representation … Semi-Supervised Representation Towards a Uniﬁed Optimization Framework for Semi-Supervised Learning Chun-Guang Li1, Zhouchen Lin2,3, Honggang

Semi Supervised e Book

Semi-Supervised Learningzhuxj/tmp/book.pdf1.1.2 Semi-Supervised Learning Semi-supervised learning (SSL) is half way between supervised and unsupervised learning. In addition to unlabeled

Augmenting Feature-driven fMRI Analyses: Semi-supervised ... · performance of feature-driven regression of stimulus-driven data. 3 Semi-supervised Learning Semi-supervised learning

Step 1. Semi-supervised

Semi-Supervised Clustering

Supervised and Semi-Supervised Multi-View Canonical Correlation … · remote sensing Article Supervised and Semi-Supervised Multi-View Canonical Correlation Analysis Ensemble for

Semi-Supervised Learning for Optical Flow with Generative … › paper › 6639-semi-supervised-learning-for-opti… · Semi-Supervised Learning for Optical Flow with Generative

S4L: Self-Supervised Semi-Supervised Learning...and existing semi-supervised learning methods. We then show that S4L and existing semi-supervised methods can be jointly trained, yielding

Incremental Semi-supervised Clustering

Semi-supervised Information Extraction

Semi-supervised Self-training for Sentence Subjectivity …ccc.inaoep.mx/~villasen/bib/Semi-supervised Self-training... · 2009-06-25 · Semi-supervised Self-training for Sentence

A semi-supervised learning algorithm via adaptive Laplacian …crabwq.github.io/pdf/2020 A Semi-supervised Learning... · 2021. 7. 31. · A semi-supervised learning algorithm via

Semi-Supervised and Self-Evolving Learning Algorithms with .../67531/metadc177238/...CHAPTER 3. SEMI-SUPERVISED LEARNING ALGORITHMS 19 3.1. Active and Semi-supervised Data Domain Description

Semi-Supervised Network Embedding

Semi-Supervised Learning - UBC Computer Scienceschmidtm/MLRG/Semi-Supervised Learning.pdf · • Semi-Supervised Learning ... Demiriz, 1999) (Demirez & Bennett, 2000) (Fung & Mangasarian,