One-shot learning of object categories

Li Fei-Fei, Robert Fergus, Pietro Perona

Research output: Contribution to journalArticle

Abstract

Learning visual models of object categories notoriously requires hundreds or thousands of training examples. We show that it is possible to learn much information about a category from just one, or a handful, of images. The key insight is that, rather than learning from scratch, one can take advantage of knowledge coming from previously learned categories, no matter how different these categories might be. We explore a Bayesian implementation of this idea. Object categories are represented by probabilistic models. Prior knowledge is represented as a probability density function on the parameters of these models. The posterior model for an object category is obtained by updating the prior in the light of one or more observations. We test a simple implementation of our algorithm on a database of 101 diverse object categories. We compare category models learned by an implementation of our Bayesian approach to models learned from by Maximum Likelihood (ML) and Maximum A Posteriori (MAP) methods. We find that on a database of more than 100 categories, the Bayesian approach produces informative models when the number of training examples is too small for other methods to operate successfully.

Original languageEnglish (US)
Pages (from-to)594-611
Number of pages18
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume28
Issue number4
DOIs
StatePublished - Apr 2006

Fingerprint

Bayesian Approach
Model Category
Probability density function
Maximum likelihood
Model
Learning
Object
Maximum a Posteriori
Prior Knowledge
Probabilistic Model
Updating
Maximum Likelihood
Training
Statistical Models
Vision
Observation
Knowledge

Keywords

  • Few images
  • Learning
  • Object categories
  • Priors
  • Recognition
  • Unsupervised
  • Variational inference

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Electrical and Electronic Engineering
  • Artificial Intelligence
  • Computer Vision and Pattern Recognition

Cite this

One-shot learning of object categories. / Fei-Fei, Li; Fergus, Robert; Perona, Pietro.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 28, No. 4, 04.2006, p. 594-611.

Research output: Contribution to journalArticle

@article{8df6afea8e074a9e9eb7834a12540209,
title = "One-shot learning of object categories",
abstract = "Learning visual models of object categories notoriously requires hundreds or thousands of training examples. We show that it is possible to learn much information about a category from just one, or a handful, of images. The key insight is that, rather than learning from scratch, one can take advantage of knowledge coming from previously learned categories, no matter how different these categories might be. We explore a Bayesian implementation of this idea. Object categories are represented by probabilistic models. Prior knowledge is represented as a probability density function on the parameters of these models. The posterior model for an object category is obtained by updating the prior in the light of one or more observations. We test a simple implementation of our algorithm on a database of 101 diverse object categories. We compare category models learned by an implementation of our Bayesian approach to models learned from by Maximum Likelihood (ML) and Maximum A Posteriori (MAP) methods. We find that on a database of more than 100 categories, the Bayesian approach produces informative models when the number of training examples is too small for other methods to operate successfully.",
keywords = "Few images, Learning, Object categories, Priors, Recognition, Unsupervised, Variational inference",
author = "Li Fei-Fei and Robert Fergus and Pietro Perona",
year = "2006",
month = "4",
doi = "10.1109/TPAMI.2006.79",
language = "English (US)",
volume = "28",
pages = "594--611",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "4",

}

TY - JOUR

T1 - One-shot learning of object categories

AU - Fei-Fei, Li

AU - Fergus, Robert

AU - Perona, Pietro

PY - 2006/4

Y1 - 2006/4

N2 - Learning visual models of object categories notoriously requires hundreds or thousands of training examples. We show that it is possible to learn much information about a category from just one, or a handful, of images. The key insight is that, rather than learning from scratch, one can take advantage of knowledge coming from previously learned categories, no matter how different these categories might be. We explore a Bayesian implementation of this idea. Object categories are represented by probabilistic models. Prior knowledge is represented as a probability density function on the parameters of these models. The posterior model for an object category is obtained by updating the prior in the light of one or more observations. We test a simple implementation of our algorithm on a database of 101 diverse object categories. We compare category models learned by an implementation of our Bayesian approach to models learned from by Maximum Likelihood (ML) and Maximum A Posteriori (MAP) methods. We find that on a database of more than 100 categories, the Bayesian approach produces informative models when the number of training examples is too small for other methods to operate successfully.

AB - Learning visual models of object categories notoriously requires hundreds or thousands of training examples. We show that it is possible to learn much information about a category from just one, or a handful, of images. The key insight is that, rather than learning from scratch, one can take advantage of knowledge coming from previously learned categories, no matter how different these categories might be. We explore a Bayesian implementation of this idea. Object categories are represented by probabilistic models. Prior knowledge is represented as a probability density function on the parameters of these models. The posterior model for an object category is obtained by updating the prior in the light of one or more observations. We test a simple implementation of our algorithm on a database of 101 diverse object categories. We compare category models learned by an implementation of our Bayesian approach to models learned from by Maximum Likelihood (ML) and Maximum A Posteriori (MAP) methods. We find that on a database of more than 100 categories, the Bayesian approach produces informative models when the number of training examples is too small for other methods to operate successfully.

KW - Few images

KW - Learning

KW - Object categories

KW - Priors

KW - Recognition

KW - Unsupervised

KW - Variational inference

UR - http://www.scopus.com/inward/record.url?scp=33144466753&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33144466753&partnerID=8YFLogxK

U2 - 10.1109/TPAMI.2006.79

DO - 10.1109/TPAMI.2006.79

M3 - Article

C2 - 16566508

AN - SCOPUS:33144466753

VL - 28

SP - 594

EP - 611

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 4

ER -