DeepLip: block-based lip pixel detection by deep neural networks

ABSTRACT

This paper presents an effective lip pixel detection method based on blocks and deep neural networks. Since only-rough localization of a pair of lips is a trivial task, we use a rectangle that loosely bounds two lips as an input region of interest for lip detection. For each pixel in the rectangle region we generate a block whose center is at the pixel, and the pixel is classified into either a lip or non-lip pixel by exploiting the pixels in the block. Deep neural networks are trained using a sufficient number of labeled blocks obtained from a quite tractable number of labeled images. As a result, lip pixels are detected with high accuracy despite negligible labeling effort. Experimental results demonstrate the effectiveness of the presented method. We show that even single-minute training can outperform the mouth map with the best threshold.

KEYWORDS:

• Lip
• detection
• block
• neural network
• lip pixel
• deeplip

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Changsoo Je received a B.S. degree in Physics and M.S. and Ph.D. degrees in Media Technology from Sogang University, Seoul, Korea, in 2000, 2002, and 2008, respectively. He is currently a Research Professor of Electronic Engineering at Sogang University. He was a Research Professor (2009-2010) and a Postdoctoral Research Fellow (2008-2009) of Computer Science and Engineering at Ewha Womans University. His research interests include computer vision, computer graphics, and image processing. He is currently serving International Journal of Sensors, Wireless Communications and Control as an Associate Board Member and Journal of Engineering and Computer Innovations (JECI) as an Editor. He received an Outstanding Paper Award at the Korea Computer Graphics Society Conference in 2008 and a Samsung Humantech Thesis Prize Award from Samsung Electronics Co., Ltd. in 2004.

Hyung-Min Park received the B.S., M.S., and Ph.D. degrees in Electrical Engineering and Computer Science from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 1997, 1999, and 2003, respectively. From 2003 to early 2005, he was a post-doc. at the Department of Biosystems, KAIST. From 2005 to early 2007, he was with the Language Technologies Institute, Carnegie Mellon University. In 2007, he joined the Department of Electronic Engineering, Sogang University, Seoul, Korea, and now is a professor. His main research interests include robust speech recognition and computer vision.

ORCID

Changsoo Je http://orcid.org/0000-0001-6153-6528

Additional information

Funding

This work was supported by the National Research Foundation (NRF) of Korea grant funded by the Korea government (MSIT) [grant number 2017R1A2B4009964].