Facial Expression 1 Facial Expression Analysis CVPR 0708

  • Slides: 16
Download presentation
Facial Expression 1. Facial Expression Analysis (CVPR’ 07&08) n Static Image based Methods n

Facial Expression 1. Facial Expression Analysis (CVPR’ 07&08) n Static Image based Methods n Video based Methods l A nature facial expression evolves over time from the onset, the apex, to the offset. l The key issue is how to describe the dynamics of the expression? National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Facial Expression 2. Facial Expression Analysis (CVPR’ 08) n Video based Methods l The

Facial Expression 2. Facial Expression Analysis (CVPR’ 08) n Video based Methods l The time-resolution issue National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Facial Expression 3. Facial Expression Analysis (CVPR’ 08) n Our Solution l Harr-like feature

Facial Expression 3. Facial Expression Analysis (CVPR’ 08) n Our Solution l Harr-like feature representation l Gaussian based temporal pattern models National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Facial Expression 4. Facial Expression Analysis (CVPR’ 08) n Our Solution l Dynamic binary

Facial Expression 4. Facial Expression Analysis (CVPR’ 08) n Our Solution l Dynamic binary patterns l Adaboost based classifier National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Facial Expression 5. Facial Expression Analysis (CVPR’ 08) National Laboratory of Pattern Recognition 模式识别国家重点实验室

Facial Expression 5. Facial Expression Analysis (CVPR’ 08) National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Facial Expression 6. Facial Expression Analysis (CVPR’ 08) National Laboratory of Pattern Recognition 模式识别国家重点实验室

Facial Expression 6. Facial Expression Analysis (CVPR’ 08) National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 1. Face Alignment (ICCV’ 07) n Active Shape Model (ASM) The original

Face Alignment 1. Face Alignment (ICCV’ 07) n Active Shape Model (ASM) The original ASM PCA based global shape model Local texture models for every shape point The issues Points are independent in search Global shape model is linear National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 2. Face Alignment (ICCV’ 07) n Component based Framework l Local model:

Face Alignment 2. Face Alignment (ICCV’ 07) n Component based Framework l Local model: PCA + Nearest neighborhood constrain between points l Global model: GPLVM National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 3. Face Alignment (ICCV’ 07) n Objective Function National Laboratory of Pattern

Face Alignment 3. Face Alignment (ICCV’ 07) n Objective Function National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 4. Face Alignment (ICCV’ 07) n MRF based Local Search National Laboratory

Face Alignment 4. Face Alignment (ICCV’ 07) n MRF based Local Search National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 5. Face Alignment (ICCV’ 07) n GPLVM based Global Constrain National Laboratory

Face Alignment 5. Face Alignment (ICCV’ 07) n GPLVM based Global Constrain National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 6. Face Alignment (ICCV’ 07) n GPLVM vs. PCA as Global Model?

Face Alignment 6. Face Alignment (ICCV’ 07) n GPLVM vs. PCA as Global Model? National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 7. Face Alignment (ICCV’ 07) n Alignment Results with Occlusion ASM PCMN

Face Alignment 7. Face Alignment (ICCV’ 07) n Alignment Results with Occlusion ASM PCMN Our method National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 8. Face Alignment (ICCV’ 07) n Alignment Results with exaggerated expression and

Face Alignment 8. Face Alignment (ICCV’ 07) n Alignment Results with exaggerated expression and off-plane rotations ASM PCMN Our method National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 9. Face Alignment (ICCV’ 07) National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute

Face Alignment 9. Face Alignment (ICCV’ 07) National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所

Face Alignment 10. Face Alignment (ICCV’ 07) n Illumination Robustness Evaluation National Laboratory of

Face Alignment 10. Face Alignment (ICCV’ 07) n Illumination Robustness Evaluation National Laboratory of Pattern Recognition 模式识别国家重点实验室 Institute of Automation, Chinese Academy of Sciences 中国科学院自动化研究所