face recognition committee machine: methodology, experiments and a system application
DESCRIPTION
Face Recognition Committee Machine: Methodology, Experiments and A System Application. Oral Defense by Sunny Tang 15 Aug 2003. Outline. Introduction Face Recognition Problems and Objectives Face Recognition Committee Machine Committee Members Result, Confidence and Weight - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/1.jpg)
Face Recognition Committee Machine:Methodology, Experiments and A System Application
Oral Defense by Sunny Tang15 Aug 2003
![Page 2: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/2.jpg)
Outline
IntroductionFace RecognitionProblems and Objectives
Face Recognition Committee Machine
Committee MembersResult, Confidence and WeightStatic and Dynamic Structure
![Page 3: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/3.jpg)
Outline
Face Recognition SystemSystem ArchitectureFace Recognition ProcessDistributed Architecture
Experimental ResultsConclusionQ & A
![Page 4: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/4.jpg)
Introduction: Face Recognition
DefinitionA recognition process that analyzes facial characteristics
Two modes of recognitionIdentification: “Who is this”Verification: “Is this person who she/he claim to be?”
![Page 5: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/5.jpg)
Face Recognition Applications
SecurityAccess control systemLaw enforcement
Multimedia databaseVideo indexingHuman search engine
![Page 6: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/6.jpg)
Problems & Objectives
Current problems of existing algorithmsNo objective comparisonAccuracy not satisfactoryCannot handle all kinds of variations
ObjectivesProvide thorough and objectively comparisonPropose a framework to integrate different algorithms for better performanceImplement a real-time face recognition system
![Page 7: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/7.jpg)
Face Recognition Committee Machine (FRCM)
MotivationAchieve better accuracy by combining predictions of different experts
Two structures of FRCMStatic structure (SFRCM)Dynamic structure (DFRCM)
![Page 8: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/8.jpg)
Static vs. Dynamic
Static structureIgnore input signalsFixed weights
Dynamic structureEmploy input signal to improve the classifiersVariable weights
![Page 9: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/9.jpg)
Committee Members
Template matching approachEigenfaceFisherfaceElastic Graph Matching (EGM)
Machine learning approachSupport Vector Machines (SVM)Neural Networks (NN)
![Page 10: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/10.jpg)
Review: Eigenface & Fisherface
Feature spaceEigenface: Principal Component Analysis (PCA)Fisherface: Fisher’s Linear Discriminant (FLD)
Training & RecognitionProject images on feature spaceCompare Euclidean distance and choose the closest projection
![Page 11: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/11.jpg)
Review: Elastic Graph Matching
Based on dynamic link architectureExtract facial feature by Gabor wavelet transformFace is represented by a graph consists of nodes of jets
Compare graphs by cost functionEdge similarity Se and vertex similarity Sv
Cost function
![Page 12: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/12.jpg)
Review: SVM & Neural Networks
SVMLook for a separating hyperplane which separates the data with the largest margin
Neural NetworksAdjust neuron weights to minimize prediction error between the target and output
![Page 13: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/13.jpg)
Result, Confidence & Weight
ResultResult of expert
ConfidenceConfidence of expert on its result
WeightWeight of expert’s result in ensemble
![Page 14: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/14.jpg)
SFRCM Architecture
![Page 15: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/15.jpg)
Result & Confidence (1)
Eigenface, Fisherface & EGMResult:
• Identification:
• Verification:
Confidence:• Identification:
• Verification:
![Page 16: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/16.jpg)
Result & Confidence (2)
SVMOne-against-one approachResult:• Identification: SVM result • Verification: direct matching
Confidence:
![Page 17: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/17.jpg)
Result & Confidence (3)
Neural networkA binary vector of size J for target representationResult:
• Identification:
• Verification:
Confidence: output value oj
![Page 18: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/18.jpg)
Weight
Derived from performance of expert:
Amplify the difference of the performance:
Normalize in range [0, 1]:
![Page 19: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/19.jpg)
Voting Machine
Assemble result and confidenceScore of expert’s result:
Ensemble result:
![Page 20: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/20.jpg)
SFRCM Drawbacks
Fixed weights under all situationsThe weights of the experts are fixed no matter which images are given.
No update mechanismThe weights cannot be updated once the system is trained
![Page 21: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/21.jpg)
DFRCM Architecture
Gating network is includedImage is involved in determination of weight
![Page 22: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/22.jpg)
Gating Network
Keep the performance of experts on different face databasesDetermine the database of input imageGive the corresponding weights of the experts for that database
![Page 23: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/23.jpg)
Feedback Mechanism1. Initialize ni,j and ti,j to 02. Train each expert i on different database j3. While TESTING
a) Determine j for each test imageb) Recognize the image in each expert ic) If ti,j != 0 then Calculate pi,j
d) Else Set pi,j = 0e) Calculate wi,j
f) Determine ensemble resultg) If FEEDBACK then Update ni,j and ti,j
4. End while
![Page 24: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/24.jpg)
Implementation: Face Recognition System
Real-time face recognition systemImplementation of FRCMFace processing
Face trackingFace detectionFace recognition
![Page 25: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/25.jpg)
System Architecture
![Page 26: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/26.jpg)
Face Recognition Process
EnrollmentCollect face images to train the experts
RecognitionIdentificationVerification
![Page 27: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/27.jpg)
System Snapshots
Identification Verification
![Page 28: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/28.jpg)
Problems of FRCM on mobile device
Memory limitationLittle memory for mobile devicesRequirement for recognition
CPU power limitationTime and storage overhead of FRCM
![Page 29: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/29.jpg)
Distributed Architecture
ClientCapture imageEnsemble results
ServerRecognition
![Page 30: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/30.jpg)
Distributed System: Evaluation
ImplementationDesktop (1400MHz), notebook (300MHz)S: Startup, R: Recognition
Distinct servers:
![Page 31: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/31.jpg)
Experimental Results
Databases used:ORL from AT&T LaboratoriesYale from Yale UniversityAR from Computer Vision Center at U.A.B HRL from Harvard Robotics Laboratory
Cross validation testing
![Page 32: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/32.jpg)
1. Apply median filter to reduce noise in background
2. Apply Sobel filter for edge detection3. Covert to a binary image4. Apply horizontal and vertical projection5. Find face boundary 6. Obtain the center of the face region.7. Crop the face region and resize it
Preprocessing
![Page 33: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/33.jpg)
ORL Result
ORL Face database
400 images40 peopleVariations
• Position• Rotation• Scale• Expression
![Page 34: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/34.jpg)
Yale Result
Yale Face Database
165 images15 peopleVariations
• Expression• Lighting
![Page 35: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/35.jpg)
AR Result
AR Face Database1300 images130 peopleVariations
• Expression• Lighting• Occlusions
![Page 36: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/36.jpg)
HRL Result
HRL Face Database
345 images5 peopleVariation
• Lighting
![Page 37: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/37.jpg)
Average Running Time & ResultsAverage running time
Average experimental results
![Page 38: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/38.jpg)
Conclusion
Make a thorough comparison of five face recognition algorithmsPropose FRCM to integrate different face recognition algorithmsImplement a face recognition system for real-time applicationPropose a distributed architecture for mobile device
![Page 39: Face Recognition Committee Machine: Methodology, Experiments and A System Application](https://reader035.vdocument.in/reader035/viewer/2022062222/56815740550346895dc4e512/html5/thumbnails/39.jpg)
Question & Answer Section
Thanks