TY - GEN
T1 - Real-Time Facial Feature Extraction Scheme Using Cascaded Networks
AU - Kim, Hyeonwoo
AU - Kim, Hyungjoon
AU - Hwang, Eenjun
N1 - Funding Information:
ACKNOWLEDGMENT This work was partly supported by Institute for Information & communications Technology Promotion (IITP) grant funded by the Korea government (MSIT) (No. R0190- 16-2012, High Performance Big Data Analytics Platform Performance Acceleration Technologies Development) and Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2016R1D1A1A09919590).
Publisher Copyright:
© 2019 IEEE.
PY - 2019/4/1
Y1 - 2019/4/1
N2 - Facial landmarks such as eyes, nose, and mouth are the most prominent feature points on the face. So far, many works have been done for efficiently extracting such landmarks from facial images. Utilizing more feature points for landmark extraction usually requires more processing time, which has been an obstacle to real-time processing or video processing. On the contrary, utilizing a too small number of feature points cannot represent diverse landmark properties such as shape accurately. In this paper, we propose a deep learning based method for extracting popular 68 feature points for facial landmarks quickly and accurately. To do that, we first detect all the faces in the image by using a cascaded structure composed of relatively light Convolution Neural Networks(CNN). Then, we perform facial landmark extraction for each face, which reduces the processing time a lot. We performed several experiments to evaluate the performance of our method. We report some of the results.
AB - Facial landmarks such as eyes, nose, and mouth are the most prominent feature points on the face. So far, many works have been done for efficiently extracting such landmarks from facial images. Utilizing more feature points for landmark extraction usually requires more processing time, which has been an obstacle to real-time processing or video processing. On the contrary, utilizing a too small number of feature points cannot represent diverse landmark properties such as shape accurately. In this paper, we propose a deep learning based method for extracting popular 68 feature points for facial landmarks quickly and accurately. To do that, we first detect all the faces in the image by using a cascaded structure composed of relatively light Convolution Neural Networks(CNN). Then, we perform facial landmark extraction for each face, which reduces the processing time a lot. We performed several experiments to evaluate the performance of our method. We report some of the results.
KW - Facial landmarks
KW - MTCNN
KW - cascaded structure
KW - face alignment
KW - face detection
KW - real-time extraction
UR - http://www.scopus.com/inward/record.url?scp=85064597565&partnerID=8YFLogxK
U2 - 10.1109/BIGCOMP.2019.8679316
DO - 10.1109/BIGCOMP.2019.8679316
M3 - Conference contribution
AN - SCOPUS:85064597565
T3 - 2019 IEEE International Conference on Big Data and Smart Computing, BigComp 2019 - Proceedings
BT - 2019 IEEE International Conference on Big Data and Smart Computing, BigComp 2019 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2019 IEEE International Conference on Big Data and Smart Computing, BigComp 2019
Y2 - 27 February 2019 through 2 March 2019
ER -