search for


Deep neural networks for lying down face detection
Journal of the Korean Data & Information Science Society 2024;35:63-73
Published online January 31, 2024;
© 2024 Korean Data and Information Science Society.

Sungho Park1 · Heewon Kim2

12Global School of Media, Soong-Sil University
Correspondence to: This paper is based on a team project (Team members: Byungkwan Chae and Jia Kim) of a class at Soongsil University.
1 Undergraduate student, Global School of Media, Soongsil University, Seoul 06978, Korea.
2 Corresponding author: Assistant professor, Global School of Media, Soongsil University, Seoul 06978, Korea. E-mail:
Received December 13, 2023; Revised January 12, 2024; Accepted January 13, 2024.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
We introduced a neural network to detect a lying-down person from facial images. We constructed a new dataset through YouTube videos, obtaining 819 labeled facial images from 42 different video clips. We fine-tuned a pre-trained Swin Transformer using this dataset for binary classification. Our model achieves 96.3% accuracy on the proposed test set and 72.7% recall rate on an existing large-scale face dataset. In addition, we developed a real-time application based on the proposed approach to evaluate its accuracy in real-world scenarios. The results of this paper have the potential to find applications across various domains, such as detecting a user’s wakefulness in smartphone alarm applications or monitoring the user’s state while wearing a virtual reality device, solely using facial images.
Keywords : Computer vision, deep learning, face recognition, image classification.