masked correlation filters for
play

Masked Correlation Filters for Partially Occluded Face Recognition - PowerPoint PPT Presentation

Masked Correlation Filters for Partially Occluded Face Recognition Eric He ICASSP 2016 2/24/2015 Partial Occlusions of Faces 2 Correlation Filters FFT IFFT Correlation Test Image Filter Training Images Filter Design T. Sim, S. Baker,


  1. Masked Correlation Filters for Partially Occluded Face Recognition Eric He ICASSP 2016 2/24/2015

  2. Partial Occlusions of Faces 2

  3. Correlation Filters FFT IFFT Correlation Test Image Filter Training Images Filter Design T. Sim, S. Baker, and M. Bsat , “The CMU pose, illumination, and 3 expression (PIE) database,” in Automatic Face and Gesture Recognition, 2002, 2002, pp. 46 – 51.

  4. Aliasing • Aliasing is a problem which results from traditional CF formulation being designed using circular correlation Input Signal Linear Correlation Circular Correlation 4

  5. Zero-Padding Images • In order to deal with aliasing, signals can be zero-padded • Our training and testing images are zero-padded 5

  6. Zero Aliasing vs Conventional Correlation Filters • ZACFs remove aliasing by removing energy in the tail of the filter • Shown below are 2D impulse responses of a traditional CF and a ZACF Traditional CF ZACF J. A. Fernandez, V. N. Boddetti, A. Rodriguez, and B. V. K. Vijaya Kumar, 6 “Zero - aliasing correlation filters for object recognition,” Pattern Analysis and Machine Intelligence, IEEE Transactions on, 2014.

  7. Zero Aliasing Correlation Filters (ZACF) • Removes aliasing effects • Sets the tail of the template 50 100 to zero 150 • 𝑩 + ഥ 𝒊 = 𝟏 200 • 𝑩 is the IDFT matrix which 250 50 100 150 200 250 when right-multiplied by a vectorized frequency domain correlation filter, results in the tail of the filter 7

  8. Masked Correlation Filters (MCF) • 𝑩 + ഥ 𝒊 = 𝟏 • 𝑩 is the IDFT matrix which when right- multiplied by a vectorized frequency domain correlation filter, results in the zeroed regions of the filter 8

  9. CMU Pose Illumination and Expression Database • Frontal, neutral expressions • Varying illuminations • PIE-lights – 68 classes – 24 images per class – Ambient lights on • PIE-nolights – 66 classes – 21 images per class – Ambient lights off 9

  10. CMU PIE Training Sets • 3 Images used for Training • Left Illumination • Frontal Illumination • Right Illumination 10

  11. Artificial Occlusions 11

  12. Scarf Results PIE-lights PIE-nolights 12

  13. Sunglasses Results PIE-lights PIE-nolights 13

  14. AR Database • Frontal Neutral Expression • Varying Lighting • 2 Types of Occlusion – Sunglasses – Scarf 14

  15. AR Training Images • 8 Images used for Training • Frontal Neutral Expression, with varying lighting 15

  16. AR Testing Set • 2 Test Sets • Scarf Test Set: 6 Scarf Images per Subject • Sunglasses Test Set: 6 Sunglasses Images per Subject 16

  17. Scarf Results 17

  18. Sunglasses Results 18

  19. KACST Dataset • 146 Classes • Neutral expression • 7 Images per Class • 2 Types of Occlusion – Sunglasses – Scarf

  20. Training Images • 4 Images per Subject • With Shemagh • With Cap • Without Headwear

  21. KACST Testing Set • 2 Test Sets • Scarf Test Set: 4 Scarf Images per Subject • Sunglasses Test Set: 2 Sunglasses Images per Subject 21

  22. Scarf Results 22

  23. Sunglasses Results 23

  24. Conclusions • Inspired by the design of Zero Aliasing CFs, we designed Masked CFs for occlusion tolerant face recognition • We showed ZACFs perform well in the face of occlusion • MCFs perform even better than ZACFs when dealing with occlusions 24

  25. Questions? 25

Recommend


More recommend