Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita...

17

  • date post

    18-Dec-2015

  • Category

    Documents

  • view

    213

  • download

    0

TAGS:

Transcript of Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita...

Page 1: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –
Page 2: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Meet the xyz

Page 3: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Spot the xyz

Page 4: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Object categories

individualobjects

`visual’ categories

`functional’categories*

Page 5: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Part similarity

Page 6: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Importance of `mutual position’

Page 7: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Presence / Absence of Features

occlusion

Page 8: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Background clutter

Page 9: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Unsupervised learning

Page 10: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Unsupervised detector training - 1

• Highly textured neighborhoods are selected automatically• produces 100-1000 patterns per image

10

10

Page 11: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Unsupervised detector training - 2

“Pattern Space” (100+ dimensions)

Page 12: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Unsupervised detector training - 3

100-1000 images ~100 detectors

Page 13: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Parameter Estimation

• Take training images. Consider set of detectors…

• Apply detectors…..

Page 14: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Rear Views of Cars

• 200 Images (100 training, 100 testing)• Only one image per car • High-pass filtered

Page 15: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Preselected Parts

Model Foreground pdf

Sample Detection

Parts in Model

Learned ModelTest Error: 13% (5 Parts)

Page 16: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Detections of Cars

Page 17: Unsupervised Learning for Recognition Pietro Perona California Institute of Technology & Universita di Padova 11 th British Machine Vision Conference –

Background Images


Perona-Malik equation error estimates for numerical finite volume scheme

Perona-Malik equation error estimates for numerical finite volume scheme

SPHEROIDAL POPULATED STAR SYSTEMS Pietro Giannone Dipartimento di Fisica Universita’ “Sapienza” Roma SUMMARY Globular clusters and star evolution Elliptical.

SPHEROIDAL POPULATED STAR SYSTEMS Pietro Giannone Dipartimento di Fisica Universita’ “Sapienza” Roma SUMMARY Globular clusters and star evolution Elliptical.

Pietro Lookbook

Pietro Lookbook

Visual identification by signature tracking - Pattern ...Mario E. Munich, Member, IEEE, and Pietro Perona, Member, IEEE ... [16], [18], [29] for surveys) and is divided into two main

Visual identification by signature tracking - Pattern ...Mario E. Munich, Member, IEEE, and Pietro Perona, Member, IEEE ... [16], [18], [29] for surveys) and is divided into two main

Visual Recognition With Humans in the Loop Steve Branson Catherine Wah Florian Schroff Boris Babenko Serge Belongie Peter Welinder Pietro Perona ECCV 2010,

Visual Recognition With Humans in the Loop Steve Branson Catherine Wah Florian Schroff Boris Babenko Serge Belongie Peter Welinder Pietro Perona ECCV 2010,

5 Pietro Perona, PhD Emmanuel Mignot, MD, PhD , and HHS ...

5 Pietro Perona, PhD Emmanuel Mignot, MD, PhD , and HHS ...

Recognition in Terra Incognita - Foundation · RecognitioninTerra Incognita Sara Beery, Grant Van Horn, and Pietro Perona Caltech {sbeery,gvanhorn,perona}@caltech.edu Abstract. It

Recognition in Terra Incognita - Foundation · RecognitioninTerra Incognita Sara Beery, Grant Van Horn, and Pietro Perona Caltech {sbeery,gvanhorn,perona}@caltech.edu Abstract. It

Effective Communication Former Congressman Mentors … · 2017. 10. 6. · Athanassios (Thanos) Siapas, Pietro Perona, Christof Koch, and Carver Mead at the 30-year celebration for

Effective Communication Former Congressman Mentors … · 2017. 10. 6. · Athanassios (Thanos) Siapas, Pietro Perona, Christof Koch, and Carver Mead at the 30-year celebration for

(Infinitely) Deep Learning in Vision Max Welling (UCI) collaborators: Ian Porteous (UCI) Evgeniy Bart UCI/Caltech) Pietro Perona (Caltech)

(Infinitely) Deep Learning in Vision Max Welling (UCI) collaborators: Ian Porteous (UCI) Evgeniy Bart UCI/Caltech) Pietro Perona (Caltech)

Mult-View Image Compositions - CaltechAUTHORS · 2012. 12. 26. · Mult-View Image Compositions Lihi Zelnik-Manor and Pietro Perona California Institute of Technology Pasadena, CA,

Mult-View Image Compositions - CaltechAUTHORS · 2012. 12. 26. · Mult-View Image Compositions Lihi Zelnik-Manor and Pietro Perona California Institute of Technology Pasadena, CA,

Object class recognition using unsupervised scale-invariant learning Rob Fergus Pietro Perona Andrew Zisserman Oxford University California Institute of.

Object class recognition using unsupervised scale-invariant learning Rob Fergus Pietro Perona Andrew Zisserman Oxford University California Institute of.

SATUAN EKSPRESI PENGUNGKAP WARNA PERONA BIBIR …

SATUAN EKSPRESI PENGUNGKAP WARNA PERONA BIBIR …

First Aid Christophe Perona

First Aid Christophe Perona

CS6670: Computer Vision - Cornell University• Fei-Fei & Perona, 2005 Interest point detector • Csurka et al. 2004 • Fei-Fei & Perona, 2005 • Sivic et al. 2005 Other methods

CS6670: Computer Vision - Cornell University• Fei-Fei & Perona, 2005 Interest point detector • Csurka et al. 2004 • Fei-Fei & Perona, 2005 • Sivic et al. 2005 Other methods

PIETRO BERETTA 3 Manuals/Manual...PIETRO BERETTA ENERClE:05JM PIETRO BERETTA PIETRO BERETTA AROONE V PIETRO BERETTA PIETRO BERETTA ENERClE:05JM Title MANUAL Beretta 92FS Spring 02AUG16

PIETRO BERETTA 3 Manuals/Manual...PIETRO BERETTA ENERClE:05JM PIETRO BERETTA PIETRO BERETTA AROONE V PIETRO BERETTA PIETRO BERETTA ENERClE:05JM Title MANUAL Beretta 92FS Spring 02AUG16

Unsupervised Learning of Visual Taxonomies IEEE conference on CVPR 2008 Evgeniy Bart – Caltech Ian Porteous – UC Irvine Pietro Perona – Caltech Max Welling.

Unsupervised Learning of Visual Taxonomies IEEE conference on CVPR 2008 Evgeniy Bart – Caltech Ian Porteous – UC Irvine Pietro Perona – Caltech Max Welling.

Fabbrica d’Armi Pietro Beretta S.p.A. Via Pietro Beretta ... · Fabbrica d’Armi Pietro Beretta S.p.A. Via Pietro Beretta, 18 - Casella Postale 160 25063 Gardone Val Trompia (BS)

Fabbrica d’Armi Pietro Beretta S.p.A. Via Pietro Beretta ... · Fabbrica d’Armi Pietro Beretta S.p.A. Via Pietro Beretta, 18 - Casella Postale 160 25063 Gardone Val Trompia (BS)

Descriptors II - courses.cs.washington.edu• Fei-Fei & Perona, 2005 • Interest point detector • Csurka et al. 2004 • Fei-Fei & Perona, 2005 • Sivic et al. 2005 • Other methods

Descriptors II - courses.cs.washington.edu• Fei-Fei & Perona, 2005 • Interest point detector • Csurka et al. 2004 • Fei-Fei & Perona, 2005 • Sivic et al. 2005 • Other methods

PERONA Alberto M.: Ensayos Sobre Video Documental y Cine

PERONA Alberto M.: Ensayos Sobre Video Documental y Cine

Scale-space and edge detection using anisotropic diffusion ... · Scale-Space and Edge Detection Using Anisotropic Diffusion PIETRO PERONA AND JITENDRA MALIK Abstracf-The scale-space

Scale-space and edge detection using anisotropic diffusion ... · Scale-Space and Edge Detection Using Anisotropic Diffusion PIETRO PERONA AND JITENDRA MALIK Abstracf-The scale-space