Face Recognition in Low Resolution Imageshji/cs519_slides/Face Recognition in Low... · Dwayne...
Transcript of Face Recognition in Low Resolution Imageshji/cs519_slides/Face Recognition in Low... · Dwayne...
Face Recognition
in
Low Resolution Images
Trey AmadorScott MatsumuraMatt Yiyang Yan
Introduction
❖ Purpose: low resolution facial recognition➢ Extract
image/video from source
➢ Identify the person in real time given a trained-database
taken from https://github.com/alexjc/neural-enhance
Face Recognition
Libraries
histogram of oriented gradients (HOG)
dlibSupport VectorMachines (SVM)
Process
● Neural Enhance library○ increase the resolution
of low pixel density○ Theano (neural network)
→ Lasagne (train) →upsampled image
● dlib○ Histogram of oriented
gradients (HOG)○ SVM○ feature descriptor for
detecting faces
→
Database
● IMDb○ Internet Movie Database is an online database of
information related to films, television programs and videogames
○ low and high resolution versions of the same image○ high-resolution 'base' image to train the Support Vector
Machine (SVM)
Support Vector Machine
for
Face Recognition
Arnold Schwarzenegger
SVM
Identify the Rock
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
SVM ++
+
+
+
-
-
-
--
+ The Rock- not The Rock
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
SVM ++
+
+
+
-
-
-
--
Separate data
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
SVM ++
+
+
+
-
-
-
--
Which line?
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
SVM ++
+
+
+
-
-
-
--
Thickest line
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
SVM +
-
-
+
+
+ +
+
- -
-
-
-
Separate data?
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
SVM +
-
-
+
+
+ +
+
- -
-
-
-
Non-linear separation
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
Generative
Adversarial Network
for
Upsampling Images
GAN
Back with The Rock
Imagessimilar to Dwayne Johnson
Image ofDwayne Johnson
Image ofDwayne Johnson
Imagessimilar to Dwayne Johnson
GAN
Generate this image?
+
Generative Network
produce an imageDiscriminative Network
real or fake
vs
How to train your
Generative
Adversarial Network
Discriminative
Network
real
fake
GAN
Train discriminative
network
Discriminative
Network
Generative
Network
random noise
backpropagation
GAN
Train both networks
negative gradient
positive gradient
Fake
Discriminative
Network
Generative
Network
random noise
backpropagation
GAN
Eventually?
Real
Discriminative
Network
Generative
Network
GAN
Upsampled
Real
Code
can be found at:
https://github.com/PresidentDwayneCamacho/super-res-face
super resolution
video samples
face recognition in enhanced-resolution video
super resolution image enhancement
boring Bruce Springsteen
100 x 100
enhanced Bruce Springsteen
200 x 200
actual Bruce Springsteen
high res
super resolution face recognition
unrecognized Bruce Springsteen
100 x 100
that’s Bruce Springsteen!
200 x 200
experimental paradigm
true face false face
high res high res
low res low res
enhanced res enhanced res
future directions
❖ find robust metric with which to filter data
❖ test efficacy of various algorithms
❖ generate larger dataset
References
[1] W. Zhao, et al. “Face Recognition: A Literature Survey.” ACM Computing Surveys, vol. 35, pp. 399-458, Dec. 2003.
[2] S.C. Park, M.K. Park, and M.G. Kang. “Super-Resolution Image Reconstruction: A Technical Overview.” IEEE Signal Processing Magazine. May
2003.
[3] D. Glasner, S. Bagon, and M. Irani. “Super-Resolution from a Single Image,” in IEEE 12th ICCV, 2009, pp 349-356.
[4] W.W. Zou and P.C. Yuen. “Very Low Resolution Face Recognition Problem.” IEEE Transactions on Image Processing, vol. 21, pp. 327-340, July
2012.
[5] A. Geitgey, "Face Recognition," GitHub repository, [Online]. Available: https://github.com/ageitgey/face_recognition. [Accessed 29 10 2017].
[6] N. Dalal and B. Triggs. “Histogram of Oriented Gradients for Human Detection” in CVPR, 2005, pp. 1-8.
[7] P. Felzenszwalb, et al. “Object Detection with Discriminantly Trained Part Based Models.” IEEE Transactions on Pattern Analysis and Machine
Intelligence, vol. 32, pp. 1627-1645, Sept. 2010.
[8] C. Cortes and V. Vladimir, "Support-Vector Networks," Machine Learning, vol. 20, no. 3, pp. 273-297, 1995.
[9] A. J. Champandard, "Neural Enhance," GitHub repository, [Online]. Available: https://github.com/alexjc/neural-enhance. [Accessed 29 10 2017].
[10] D. G. Lowe, "Object Recognition from Local Scale-Invariant Features," Computer Vision, vol. 2, pp. 1150-1157, 1999.
[11] K. Simonyan, M. O. Parkhi, A. Vedaldi and A. Zisserman, "Fisher Vector Faces in the Wild," British Machine Vision Conference, vol. 2, no. 3, p. 4,
Sept. 2013.
[12] P. Fischer, A. Dosovitskiy and T. Brox, "Descriptor Matching with Convolutional Neural Networks: a Comparison to SIFT," arXiv, p. 10, 22 May
2014.
[13] M. O. Parkhi, A. Vedaldi and A. Zisserman, "Deep Face Recognition," British Machine Vision Conference, vol. 1, no. 3, p. 6, 2015.
[14] U. Karn, "An Intuitive Explanation of Convolutional Neural Networks," The Data Science Blog, [Online]. Available:
https://ujjwalkarn.me/2016/08/11/intuitive-explanation-convnets/. [Accessed 29 10 2017].
[15] C. Ledig, et al. "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network," arXiv, p. 19, 25 May 2016.
[16] A.V. Nefian. “Georgia Tech Face Database.” Nov. 15, 1999. [Online]. Available: www.anefian.com/research/face_reco.htm. [Accessed: Nov. 5,
2017].
[17] Y.D. Wong. “ChokePoint Dataset.” [Online]. Available: arma.sourceforge.net/chokepoint/. [Accessed: Nov. 5, 2017].