2021年2月25日 星期四

【課程簡報】Edge AI社聚#1 身份辨識防疫技術交流 ─ Edge AI人臉辨識技術剖析

很高興今日(2021/2/25)受邀在「Edge AI 社聚」和大家分享關於邊緣智能上人臉辨識相關技術。此次從傳統算法介紹到目前人工智慧算法,並分別解釋相關辨識基準,主要包括下列項目,希望能幫助大家能快速進入「人臉辨識」這個領域

  • 人臉辨識應用
  • 人臉辨識技術
  • 人臉辨識發展
  • 人臉資料集
  • 人臉辨識基準
  • 開源辨識工具
另外文末附上14個常見人臉特徵點(Facial Landmarks)資料集載點,可自行參考。

 (以下圖點擊後放大)











 














 
影片來源:Youtube(如遭刪除敬請見諒)

 

人臉特徵點資料集:


MTFL, 12,955 images with 5 landmarks
http://mmlab.ie.cuhk.edu.hk/projects/TCDCN.html

FRGC-V2, with 5 landmarks
https://www.nist.gov/programs-projects/face-recognition-grand-challenge-frgc

AFW, 205 images with 6 landmarks
https://www.cs.cmu.edu/~deva/papers/face/

BioID Face, 1,521 images with 20 landmarks
https://www.bioid.com/facedb/

AFLW, 25,933 images with 21 landmarks
https://www.tugraz.at/institute/icg/research/team-bischof/lrs/downloads/aflw/

COFW, 1,345 images with 29 landmarks
http://www.vision.caltech.edu/xpburgos/ICCV13/

LFPW, 1,432 images with 35 landmarks
https://neerajkumar.org/databases/lfpw/

BUHMAP-DB, 440 videos with 52 landmarks
https://www.cmpe.boun.edu.tr/pilab/pilabfiles/databases/buhmap/

XM2VTS, 2,360 images with 68 landmarks
https://personalpages.manchester.ac.uk/staff/timothy.f.cootes/data/xm2vts/xm2vts_markup.html

300W, 3,837 images with 68 landmarks
https://ibug.doc.ic.ac.uk/resources/300-W/

MUCT, 3755 images with 76 landmarks
http://www.milbo.org/muct/

WFLW, 10,000 images with 98 landmarks
https://wywu.github.io/projects/LAB/WFLW.html

ICME, 16,000 images with 106 landmarks
https://facial-landmarks-localization-challenge.github.io/#introduction

HELEN, 2,330 images with 194 landmarks
http://www.ifp.illinois.edu/~vuongle2/helen/
 

參考文獻

National Institute of Standards and Technology (NIST), Face Recognition Vendor Test (FRVT) Ongoing
https://www.nist.gov/programs-projects/face-recognition-vendor-test-frvt-ongoing


許哲豪,【開箱測試】研華科技AI人臉辨識運算智能系統
http://omnixri.blogspot.com/2020/12/ai.html 

OpenCV Tutorials for face module (4.0.0)
https://docs.opencv.org/4.0.0/de/d27/tutorial_table_of_content_face.html

Overview of OpenVINO™ Toolkit Intel's Pre-Trained Models
https://docs.openvinotoolkit.org/latest/omz_models_intel_index.html

Overview of OpenVINO™ Toolkit Public Models
https://docs.openvinotoolkit.org/latest/omz_models_public_index.html


1 則留言:

【頂置】簡報、源碼、系列文快速連結區

常有人反應用手機瀏覽本部落格時常要捲很多頁才能找到系列發文、開源專案、課程及活動簡報,為了方便大家快速查詢,特整理連結如下,敬請參考! Edge AI Taiwan 邊緣智能交流區 全像顯示與互動交流區 台科大(NTUST) 人工智慧與邊緣運算實務 開南大學...