SlideShare une entreprise Scribd logo
1  sur  46
Télécharger pour lire hors ligne
A B I N J O: U C 
 L  C R  I

  Hannes Schulz, Hauke Strasdat, and Sven Behnke

                    University of Freiburg
               Institute of Computer Science



                    Nov 29, 2007
Balls and Non-Balls        Finding Ball Candidates   Classification of Ball Candidates   Experiments


 M




       The ball

               is small,
Balls and Non-Balls        Finding Ball Candidates   Classification of Ball Candidates   Experiments


 M




       The ball

               is small,
               is easy to confuse with other objects
Balls and Non-Balls        Finding Ball Candidates   Classification of Ball Candidates   Experiments


 M




       The ball

               is small,
               is easy to confuse with other objects
               is the most important object on the field:
               You cannot play sensibly without knowing its position
Balls and Non-Balls        Finding Ball Candidates   Classification of Ball Candidates   Experiments


 M




       The ball

               is small,
               is easy to confuse with other objects
               is the most important object on the field:
               You cannot play sensibly without knowing its position

          We should put a lot of effort into finding the single real ball.
O



  1   I  B  N-B


  2   F B C


  3   C  B C


  4   E
O



  1   I  B  N-B


  2   F B C


  3   C  B C


  4   E
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D B L L?
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D B L L?



                                    T E C
Balls and Non-Balls   Finding Ball Candidates    Classification of Ball Candidates   Experiments


 H D B L L?



                                    L C
Balls and Non-Balls   Finding Ball Candidates     Classification of Ball Candidates   Experiments


 H D B L L?



                                    M B
Balls and Non-Balls   Finding Ball Candidates    Classification of Ball Candidates   Experiments


 H D B L L?



                                    C W L
Balls and Non-Balls   Finding Ball Candidates      Classification of Ball Candidates   Experiments


 H D B L L?



                                    C R
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D N-B L L?
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D N-B L L?



                                    H  F
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D N-B L L?



                                    O O
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D N-B L L?



                                    O O
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D N-B L L?



                                    F
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H D N-B L L?



                                    F
O



  1   I  B  N-B


  2   F B C


  3   C  B C


  4   E
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 C  E  YUV S
Balls and Non-Balls   Finding Ball Candidates     Classification of Ball Candidates   Experiments


 C  E  YUV S




                                                Two ellipses for “orange”
                                                      Actual ball color
                                                      Wider, brownish color
Balls and Non-Balls   Finding Ball Candidates     Classification of Ball Candidates   Experiments


 C  E  YUV S




                                                Two ellipses for “orange”
                                                      Actual ball color
                                                      Wider, brownish color
                                                      Allows for motion blur
Balls and Non-Balls   Finding Ball Candidates            Classification of Ball Candidates     Experiments


 S  C I



                                                                               orange
           YUV Camera Image
                                                                               (-candidate)


                                                                               white


                                                                               green

                                                64 : 1
Balls and Non-Balls   Finding Ball Candidates            Classification of Ball Candidates     Experiments


 S  C I


           1. Find Maximum
           2. Find Weighted Mean                                               orange
                                                                               (-candidate)


                                                                               white


                                                                               green

                                                64 : 1
Balls and Non-Balls   Finding Ball Candidates            Classification of Ball Candidates     Experiments


 S  C I



                                                                               orange
        cut corresponding area
                                                                               (-candidate)


                                                                               white


                                                                               green

                                                64 : 1
Balls and Non-Balls     Finding Ball Candidates            Classification of Ball Candidates     Experiments


 S  C I



                                                                                 orange
        cut corresponding area
                                                                                 (-candidate)


                                                                                 white


       Box size depends on position in image
                                                                                 green

                                                  64 : 1
O



  1   I  B  N-B


  2   F B C


  3   C  B C


  4   E
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 P  C
Balls and Non-Balls      Finding Ball Candidates   Classification of Ball Candidates   Experiments


 P  C




               Projection changes with ellipses: Robust to changes in
               lighting conditions
Balls and Non-Balls      Finding Ball Candidates   Classification of Ball Candidates   Experiments


 P  C




               Projection changes with ellipses: Robust to changes in
               lighting conditions
Balls and Non-Balls      Finding Ball Candidates         Classification of Ball Candidates   Experiments


 P  L



                 YUV-Image                         Y-Image                Subsampled Y-Image




               Subtraction of mean: Robust to changes in lighting conditions
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H L F
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H L F
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 H L F


                                                                         -
                                                                       - + -
                                                                         -
                                                                         -
                                                                       - + -
                                                                         -
Balls and Non-Balls   Finding Ball Candidates   Classification of Ball Candidates   Experiments


 N N C




                                                               1: Ball


                                                              0: No Ball


           -
         - + -
           -
O



  1   I  B  N-B


  2   F B C


  3   C  B C


  4   E
Balls and Non-Balls   Finding Ball Candidates              Classification of Ball Candidates   Experiments


 NR KS 06 – T

                                                H
                                                    520 MHz ARM PocketPC
                                                    VGA HTC Camera

                                                D S
                                                    160 balls
                                                    440 non-balls
                                                    divided randomly in training set
                                                    (80%) and test set (20%)
Balls and Non-Balls   Finding Ball Candidates              Classification of Ball Candidates   Experiments


 NR KS 06 – T

                                                H
                                                    520 MHz ARM PocketPC
                                                    VGA HTC Camera

                                                D S
                                                    160 balls
                                                    440 non-balls
                                                    divided randomly in training set
                                                    (80%) and test set (20%)

                                                P (T S)
                                                    100% of distractors classified
                                                    correctly
                                                    1 ball out of 32 not recognized
Balls and Non-Balls   Finding Ball Candidates           Classification of Ball Candidates   Experiments


 NR KS 06 – R T


                                                H
                                                    520 MHz ARM PocketPC
                                                    VGA HTC Camera

                                                R T T




                                                    Robot decides autonomously
                                                    which object to approach.
Balls and Non-Balls   Finding Ball Candidates    Classification of Ball Candidates   Experiments


 NR KS 2007

                                                H
                                                     1.33 GHz PC
                                                     WVGA µEye Camera
Balls and Non-Balls   Finding Ball Candidates              Classification of Ball Candidates   Experiments


 NR KS 2007

   S D                                  D S
                                                    273 balls
                                                    548 non-balls
                                                    training set 62%, validation set
                                                    13%, test set 25%
                                                    varying lighting conditions
Balls and Non-Balls   Finding Ball Candidates              Classification of Ball Candidates   Experiments


 NR KS 2007

   A S                             D S
        Avg Luminance                               273 balls
   ball            non-ball                         548 non-balls
                                                    training set 62%, validation set
                                                    13%, test set 25%
                                                    varying lighting conditions

    Avg Orange-Greenness
   ball            non-ball
Balls and Non-Balls   Finding Ball Candidates              Classification of Ball Candidates   Experiments


 NR KS 2007

   A S                             D S
        Avg Luminance                               273 balls
   ball            non-ball                         548 non-balls
                                                    training set 62%, validation set
                                                    13%, test set 25%
                                                    varying lighting conditions

    Avg Orange-Greenness                        R  T S
   ball            non-ball                         91.1% accuracy
                                                    76.6% if stimuli flipped up/down.
                                                    Drop suggests dependency on
                                                    gradient.
                                                    88.2% if lighting in testset differs.
                                                    Classifier seems indifferent.
Balls and Non-Balls       Finding Ball Candidates          Classification of Ball Candidates   Experiments


 C  C




                                               Classifier
       Task           Neural Net        Linear Classifier            KNN (k = 5)
       Regular            91.1%                     86.5%                      88.5%
       Flipped            74.0%                     70.0%                      76.6%
Balls and Non-Balls      Finding Ball Candidates   Classification of Ball Candidates   Experiments


 C




               The ball has properties aside from being orange
               These properties are exploited by our Neural Network
               Classifiers
               Changes in Lighting conditions can be dealt with by projection
               to lines in YUV-space and Luminance.
               The method introduced here can be generalized to other
               small-sized objects on the field.

Contenu connexe

En vedette

Using Communication Effectively
Using Communication EffectivelyUsing Communication Effectively
Using Communication Effectively
Owen Kulemeka
 
Cascao Cortona Gambella Sharing Water Land Resources
Cascao Cortona Gambella Sharing Water Land ResourcesCascao Cortona Gambella Sharing Water Land Resources
Cascao Cortona Gambella Sharing Water Land Resources
Ana Cascao
 
Trauma Equipo7 SecciòN13
Trauma Equipo7 SecciòN13Trauma Equipo7 SecciòN13
Trauma Equipo7 SecciòN13
Adrián Chávez
 
AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...
AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...
AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...
Ana Cascao
 
Starbucks Brandz Report
Starbucks Brandz ReportStarbucks Brandz Report
Starbucks Brandz Report
pmbaird
 
13 Frases Para Vivir
13 Frases Para Vivir13 Frases Para Vivir
13 Frases Para Vivir
sutrisno2629
 
Cascao London Resistance Counter-hegemony Transboundary Basins
Cascao London Resistance Counter-hegemony Transboundary BasinsCascao London Resistance Counter-hegemony Transboundary Basins
Cascao London Resistance Counter-hegemony Transboundary Basins
Ana Cascao
 

En vedette (20)

Using Communication Effectively
Using Communication EffectivelyUsing Communication Effectively
Using Communication Effectively
 
Knovel2008
Knovel2008Knovel2008
Knovel2008
 
Coffee1
Coffee1Coffee1
Coffee1
 
Cascao Cortona Gambella Sharing Water Land Resources
Cascao Cortona Gambella Sharing Water Land ResourcesCascao Cortona Gambella Sharing Water Land Resources
Cascao Cortona Gambella Sharing Water Land Resources
 
Trauma Equipo7 SecciòN13
Trauma Equipo7 SecciòN13Trauma Equipo7 SecciòN13
Trauma Equipo7 SecciòN13
 
Para Pensar 12
Para Pensar 12Para Pensar 12
Para Pensar 12
 
Using GradeMark For Effective Feedback
Using GradeMark For Effective FeedbackUsing GradeMark For Effective Feedback
Using GradeMark For Effective Feedback
 
AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...
AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...
AEGIS Conference of African Studies 2011 - Uppsala [Land grabbing in Gambella...
 
Seconda Campagna Indagini Relazione Arta Vasto
Seconda Campagna Indagini Relazione Arta VastoSeconda Campagna Indagini Relazione Arta Vasto
Seconda Campagna Indagini Relazione Arta Vasto
 
Lead Management
Lead ManagementLead Management
Lead Management
 
The Changing Landscape of Public Relations
The Changing Landscape of Public RelationsThe Changing Landscape of Public Relations
The Changing Landscape of Public Relations
 
Starbucks Brandz Report
Starbucks Brandz ReportStarbucks Brandz Report
Starbucks Brandz Report
 
Dades i operadors
Dades i operadorsDades i operadors
Dades i operadors
 
13 Frases Para Vivir
13 Frases Para Vivir13 Frases Para Vivir
13 Frases Para Vivir
 
Renkomäen koulun veso 30.5.2011
Renkomäen koulun veso 30.5.2011Renkomäen koulun veso 30.5.2011
Renkomäen koulun veso 30.5.2011
 
Tudlo journey
Tudlo journeyTudlo journey
Tudlo journey
 
Adding A Course Using Web Reg Revised
Adding A Course Using Web Reg RevisedAdding A Course Using Web Reg Revised
Adding A Course Using Web Reg Revised
 
CEx sells @ Bienalto
CEx sells @ BienaltoCEx sells @ Bienalto
CEx sells @ Bienalto
 
Cascao London Resistance Counter-hegemony Transboundary Basins
Cascao London Resistance Counter-hegemony Transboundary BasinsCascao London Resistance Counter-hegemony Transboundary Basins
Cascao London Resistance Counter-hegemony Transboundary Basins
 
Tu Amistad
Tu AmistadTu Amistad
Tu Amistad
 

Plus de cijat

Plus de cijat (7)

Learning Kinematics from direct Self-Observation using Nearest-Neighbour Methods
Learning Kinematics from direct Self-Observation using Nearest-Neighbour MethodsLearning Kinematics from direct Self-Observation using Nearest-Neighbour Methods
Learning Kinematics from direct Self-Observation using Nearest-Neighbour Methods
 
Toward Optimal Configuration Space Sampling
Toward Optimal Configuration Space SamplingToward Optimal Configuration Space Sampling
Toward Optimal Configuration Space Sampling
 
CRF-Filters: Discriminative Particle Filters for Sequential State Estimation
CRF-Filters: Discriminative Particle Filters for Sequential State EstimationCRF-Filters: Discriminative Particle Filters for Sequential State Estimation
CRF-Filters: Discriminative Particle Filters for Sequential State Estimation
 
RoboCup Introduction
RoboCup IntroductionRoboCup Introduction
RoboCup Introduction
 
Local and Global Gating of Synaptic Plasticity
Local and Global Gating of Synaptic PlasticityLocal and Global Gating of Synaptic Plasticity
Local and Global Gating of Synaptic Plasticity
 
Auxiliary
AuxiliaryAuxiliary
Auxiliary
 
Affordances
AffordancesAffordances
Affordances
 

Dernier

Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Safe Software
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
?#DUbAI#??##{{(☎️+971_581248768%)**%*]'#abortion pills for sale in dubai@
 

Dernier (20)

FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024
 
Ransomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdfRansomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdf
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
 
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
 
Navi Mumbai Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Navi Mumbai Call Girls 🥰 8617370543 Service Offer VIP Hot ModelNavi Mumbai Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Navi Mumbai Call Girls 🥰 8617370543 Service Offer VIP Hot Model
 
Apidays Singapore 2024 - Modernizing Securities Finance by Madhu Subbu
Apidays Singapore 2024 - Modernizing Securities Finance by Madhu SubbuApidays Singapore 2024 - Modernizing Securities Finance by Madhu Subbu
Apidays Singapore 2024 - Modernizing Securities Finance by Madhu Subbu
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Script
 
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challenges
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of Terraform
 

The ball is not just orange.

  • 1. A B I N J O: U C  L  C R  I Hannes Schulz, Hauke Strasdat, and Sven Behnke University of Freiburg Institute of Computer Science Nov 29, 2007
  • 2. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments M The ball is small,
  • 3. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments M The ball is small, is easy to confuse with other objects
  • 4. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments M The ball is small, is easy to confuse with other objects is the most important object on the field: You cannot play sensibly without knowing its position
  • 5. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments M The ball is small, is easy to confuse with other objects is the most important object on the field: You cannot play sensibly without knowing its position We should put a lot of effort into finding the single real ball.
  • 6. O 1 I  B  N-B 2 F B C 3 C  B C 4 E
  • 7. O 1 I  B  N-B 2 F B C 3 C  B C 4 E
  • 8. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D B L L?
  • 9. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D B L L? T E C
  • 10. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D B L L? L C
  • 11. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D B L L? M B
  • 12. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D B L L? C W L
  • 13. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D B L L? C R
  • 14. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D N-B L L?
  • 15. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D N-B L L? H  F
  • 16. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D N-B L L? O O
  • 17. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D N-B L L? O O
  • 18. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D N-B L L? F
  • 19. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H D N-B L L? F
  • 20. O 1 I  B  N-B 2 F B C 3 C  B C 4 E
  • 21. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments C  E  YUV S
  • 22. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments C  E  YUV S Two ellipses for “orange” Actual ball color Wider, brownish color
  • 23. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments C  E  YUV S Two ellipses for “orange” Actual ball color Wider, brownish color Allows for motion blur
  • 24. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments S  C I orange YUV Camera Image (-candidate) white green 64 : 1
  • 25. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments S  C I 1. Find Maximum 2. Find Weighted Mean orange (-candidate) white green 64 : 1
  • 26. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments S  C I orange cut corresponding area (-candidate) white green 64 : 1
  • 27. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments S  C I orange cut corresponding area (-candidate) white Box size depends on position in image green 64 : 1
  • 28. O 1 I  B  N-B 2 F B C 3 C  B C 4 E
  • 29. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments P  C
  • 30. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments P  C Projection changes with ellipses: Robust to changes in lighting conditions
  • 31. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments P  C Projection changes with ellipses: Robust to changes in lighting conditions
  • 32. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments P  L YUV-Image Y-Image Subsampled Y-Image Subtraction of mean: Robust to changes in lighting conditions
  • 33. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H L F
  • 34. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H L F
  • 35. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments H L F - - + - - - - + - -
  • 36. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments N N C 1: Ball 0: No Ball - - + - -
  • 37. O 1 I  B  N-B 2 F B C 3 C  B C 4 E
  • 38. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments NR KS 06 – T H 520 MHz ARM PocketPC VGA HTC Camera D S 160 balls 440 non-balls divided randomly in training set (80%) and test set (20%)
  • 39. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments NR KS 06 – T H 520 MHz ARM PocketPC VGA HTC Camera D S 160 balls 440 non-balls divided randomly in training set (80%) and test set (20%) P (T S) 100% of distractors classified correctly 1 ball out of 32 not recognized
  • 40. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments NR KS 06 – R T H 520 MHz ARM PocketPC VGA HTC Camera R T T Robot decides autonomously which object to approach.
  • 41. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments NR KS 2007 H 1.33 GHz PC WVGA µEye Camera
  • 42. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments NR KS 2007 S D D S 273 balls 548 non-balls training set 62%, validation set 13%, test set 25% varying lighting conditions
  • 43. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments NR KS 2007 A S D S Avg Luminance 273 balls ball non-ball 548 non-balls training set 62%, validation set 13%, test set 25% varying lighting conditions Avg Orange-Greenness ball non-ball
  • 44. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments NR KS 2007 A S D S Avg Luminance 273 balls ball non-ball 548 non-balls training set 62%, validation set 13%, test set 25% varying lighting conditions Avg Orange-Greenness R  T S ball non-ball 91.1% accuracy 76.6% if stimuli flipped up/down. Drop suggests dependency on gradient. 88.2% if lighting in testset differs. Classifier seems indifferent.
  • 45. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments C  C Classifier Task Neural Net Linear Classifier KNN (k = 5) Regular 91.1% 86.5% 88.5% Flipped 74.0% 70.0% 76.6%
  • 46. Balls and Non-Balls Finding Ball Candidates Classification of Ball Candidates Experiments C The ball has properties aside from being orange These properties are exploited by our Neural Network Classifiers Changes in Lighting conditions can be dealt with by projection to lines in YUV-space and Luminance. The method introduced here can be generalized to other small-sized objects on the field.