Face Detection

  1. Current Most Frontier
  2. Reference
  3. Ensembles – Bagging and Boosting

  4. Trained Haar Cascades
  5. Modesto Castrillón-Santana‘s Haar Cascade online resources.
    Alejandro Reimondo‘s Haar Cascade online resources.
    You are also welcome to try cascadePUT.xml, which is trained by me using the face database PUT.

  6. Face Detection Results
  7. Based on those well-trained Haar/Haar-like and LBP cascade classifiers in OpenCV 2.4.2, face detection experimental results are summarized as follows:

    Database

    PUT

    BioID

    FRANCK

    XM2VTS
    (CDS001+006)

    IMM

    EMOUNT

    JIAPei

    Overall

    No. Of Images

    9,971

    1,521

    5,000

    2,360

    240

    268

    67

    19,427

    No. Of Unused Images

    0

    18

    0

    3

    0

    0

    0

    21

    Undetected by Haar

    1,650

    37

    0

    28

    0

    0

    0

    1,715

    Undetected by LBP

    284

    92

    0

    16

    0

    0

    0

    392

    DEye>0.5 Haar

    1,723

    66

    7

    47

    3

    1

    0

    1,847

    DEye>0.5 LBP

    360

    102

    1

    25

    0

    1

    0

    489

    Error Rate by Haar

    17.28%

    4.39%

    0.14%

    1.99%

    1.25%

    0.37%

    0

    9.52%

    Error Rate by LBP

    3.61%

    6.79%

    0.02%

    1.06%

    0

    0.37%

    0

    2.52%

    • Note: Please refer to chapter 3.7.1 in [1] for the definition of DEye.
  8. Face Components Detection Results
  9. After LBP based boosting cascade face detection, by confining the face components in some pre-defined rectangles, experimental results of Haar & Haar-like feature based face component detection are summarized as follows:

    Database

    PUT

    BioID

    FRANCK

    XM2VTS
    (CDS001+006)

    IMM

    EMOUNT

    JIAPei

    Overall

    Error Rate

    No. Of Tested Images

    9,971

    1,503

    5,000

    2,357

    240

    268

    67

    19,406

     

    Undetected Face by LBP

    284

    92

    0

    16

    0

    0

    0

    392

    2.02%

    Undetected Left Eye

    1,569

    527

    32

    606

    19

    18

    28

    2,799

    14.42%

    Undetected Right Eye

    2,399

    604

    37

    695

    23

    22

    42

    3,822

    19.70%

    Undetected Nose

    1,819

    262

    2

    107

    32

    5

    0

    2,227

    11.48%

    Undetected Mouth

    639

    124

    5

    79

    0

    1

    0

    848

    4.37%

    Video Effects

    PUT

    BioID

    FRANCK

    XM2VTS

    IMM

    Emount

    JIAPei

       

    Bibliography

    1. David Cristinacce. Automatic Detection of Facial Features in Grey Scale Images. PhD thesis, University of Manchester, 2004.

     

In Archive

Printed from: http://visionopen.com/resources/computer-vision/face-detection/

Scan to visit this page:

Welcome

Log in

Lost your password?

Register For This Site

Join

Join the forum!