3734378

Embed Size (px)

Citation preview

  • 8/3/2019 3734378

    1/23

    Automated Face Detection

    and RecognitionA Survey

    Waldir [email protected]

    Universidade do MinhoMestrado em Informtica

    MI-STAR 2010

  • 8/3/2019 3734378

    2/23

    Face Detection

    Locating generic faces in images

    2009 Angelo State University

  • 8/3/2019 3734378

    3/23

    Face Detection: applications

    Web cams that track the user Cameras that shoot automatically when they

    detect smiles Blurring of faces in

    public image databases

    2009 Google

    Counting of people in aroom (e.g. fortemperature adjustment)

  • 8/3/2019 3734378

    4/23

    Face Recognition

    Distinguishing a specific face from other faces

    2009 TotallyLooksLike.com

  • 8/3/2019 3734378

    5/23

    Face Recognition: applications

    Biometrics / access control

    ""Minority Report" 2002 20th Century Fox

    Superbad" 2007 Columbia Pictures

    Searching mugshotdatabases

    Tagging photo albums Detecting fake ID cards

    o no action requiredo scan many people at onceo places: airports, banks, safeso data: laptops, medical info

  • 8/3/2019 3734378

    6/23

    Humans vs. Computers

    "Built-in" face detection /recognition ability

    detection &recognition in

    different areasof the brain can be fooled

    by look-alikes

    SingularityHub.com

    Algorithms must bebuilt from scratch Virtually perfect

    memory

    Can work 24/7without degradingperformance

    Can apply stricter

    matching criteria

  • 8/3/2019 3734378

    7/23

    Computer representation of faces

    Faces vary across many attributes they'remultidimensional

    Plotted in spaces with more than 3 dimensionso in fact, it's commonly one dimension per pixelo on a 2020px image, that's 400 dimensions!

    Humans can't visualize or compute distancesintuitively in >3D space. Computers can. But...

    It is computationally intensive. Dimensionality

    reduction is applied to enhance efficiency

  • 8/3/2019 3734378

    8/23

    PCA: Principal component analysis

    Data is projected into a lower dimensional space preserving the directions that are most significant not necessarily orthogonal to the original ones!

    cc-by Lydia E. Kavraki

  • 8/3/2019 3734378

    9/23

    What defines a "match"?

    Ideally, distance in "facespace" should be:o zero, for a specific match in face recognitiono small, for a generic faceo large, otherwise

    But there are variations due to:o facial expressionso illumination varianceo pose (orientation)o dimensionality reduction

  • 8/3/2019 3734378

    10/23

    The distance theshold

    faces closer to each other than a given limit(threshold) are considered matches.

    A looser threshold can be used for facedetection.

    1991 M. Turk and A. Pentland

  • 8/3/2019 3734378

    11/23

    The ROC curve

    Too low threshold = more false negatives Too high threshold = more false positives

    EER = Equal error rate

    2007 Y. Du and C.-I. Chang"Handbook of Fingerprint Recognition" 2004 D. Maltoni et al.

  • 8/3/2019 3734378

    12/23

    Some history...

    Francis Galton (1888)Designed a biometric system for descriptionand identification of faces

    2007 University of Texas at Austin

    Public Domain

    Woody Bledsoe (1964)First implementation of automatic facial

    recognition in a mug shot database.

    Michael D. Kelly (1970)o Visual identification of people by computer

    Takeo Kanade (1973)o Computer recognition of human faces

  • 8/3/2019 3734378

    13/23

    ClassificationZhao et al., 2003:[The facial recognition problem has] attracted researchers from very

    diverse backgrounds: psychology, pattern recognition, neural

    networks, computer vision, and computer graphics.geometric (feature based) photometric (image based)

    detection recognitionpre-processing

    3DVideo

  • 8/3/2019 3734378

    14/23

    Pre-processing Face location / normalization Later processing doesn't need to scan the whole image Morphological operators (very fast) Rough operators to detect heads Finer confirmation operators to detect prominent features

    Brunelli and Poggio 1993 Reisfeld et al., 1995

    f

  • 8/3/2019 3734378

    15/23

    Eigenfaces Sirovich and Kirby 1987; Turk and Pentland 1991 Uses PCA to discover principal components (eigenvectors) Each face is described as a linear combination of the main

    eigenvectors Image-based approach

    (features might not be intuitive)

    eigenvectors can be translated backto the original pixelbasedrepresentation, many producingface-like images (hence the nameeigenfaces)

    AT&T Laboratories

    Fi h f

  • 8/3/2019 3734378

    16/23

    Fisherfaces Instead of PCA, it uses Linear disciminant analysis (LDA),

    developed by Robert Fisher in 1936 Variation can be greater due to lighting than due to different

    faces (Moses el al. 1994)

    1997 Belhumeuret al.

    Shashua [1994] demonstrated thatimages from same face but underdifferent illumination conditionslie close to each other in the high-dimensional facespace

    LDA can grasp these similarities better than PCA, whichmakes Fisherfaces more illumination independent thaneigenfaces

    N l t k

  • 8/3/2019 3734378

    17/23

    Neural networks Based on the natural brain structure

    of simple, interconnected neurons Good at approximating complex prob-

    lems without deterministic solutions Each pixel of the face image is mapped to an input neuron The intermediate (hidden-layer) neurons are as many as the

    number of reduced dimensions that are intended. The network learns what patterns are likely faces or not Initially promising, but Cottrell and Fleming [1990] showed

    that they can at best match an eigenface approach.

    cc-by-sa Cburnett

    G b l t

  • 8/3/2019 3734378

    18/23

    Gabor wavelets

    First proposed in 1968 by Dennis Gabor

    Analog to Fourier series: images are decomposed in a seriesof wavelets applied in different points Further developed to flexible models: elastic grid matching.

    GFDL Wikimedia Commons

    Wiskott et al. 1997

    A ti Sh /A M d l

  • 8/3/2019 3734378

    19/23

    Active Shape/Appearance Models Original concept by Kass et al., 1987: snakes, deformable

    curves that adjust to edges Yuille [1987] extended the concept to flexible sets of

    geometrically related points (not necessarily on a curve) Cootes [2001] applies statistical analysis to model and

    restrict the variation (flexibility) of model points

    2001 Cootes et al.

    3D

  • 8/3/2019 3734378

    20/23

    3D 2D deal poorly with varying poses

    (orientation) of the head

    Many have attempted to compensate bystoring several views per faceo obviously resource-consuming

    3D attempts to solve this issue, using:

    2006 Bowyeret al.

    active range sensors (laser scanners, ultrasound)

    passive sensors (structured light: grid projected on face) New poses can be matched by deforming the 3D model

    Vid

  • 8/3/2019 3734378

    21/23

    Video Lower quality images (frames), due to compression.

    Reconstructed models will have low accuracy. Advantage: temporal coherence, optical flow Simplest approach: use frame difference to detect moving

    foreground objects and match their shapes (blobs) to heads Locate faces, then track them

    Reconstruct 3D shape from relativemovement of tracked points. This iscalled Structure from Motion (SfM)

    2010 Christian Rakete

    Comparison

  • 8/3/2019 3734378

    22/23

    ComparisonStandard tests needed for valid results comparison

    Databases: FERET, MIT, Yale, and many smaller onesEvaluations: Face Recognition Vendor Test (FVRT) Face Recognition Grand Challenge

    XM2VTS

    Conferences: International Conference in Audio- and Video-Based Person

    Authentication (AVBPA) International Conference in Automatic Face and Gesture

    Recognition (AFGR)

  • 8/3/2019 3734378

    23/23

    Questions?