Driver Drowsiness Detection Using a Hybrid Vision Transformer and Recurrent Neural Network Model

Authors

  • يعرب ديوب جامعةِ طرطوس – كليَّةُ هندسةِ تكنولوجيا المعلوماتِ وَ الاتِّصالاتِ – قسمُ هندسةِ تكنولوجيا المعلوماتِ
  • فاطمة قداد جامعةِ طرطوس – كليَّةُ هندسةِ تكنولوجيا المعلوماتِ وَ الاتِّصالاتِ – قسمُ هندسةِ تكنولوجيا المعلوماتِ
  • آيه عباس جامعةِ طرطوس- كليَّةُ هندسةِ تكنولوجيا المعلوماتِ والاتّصالاتِ – قسمُ هندسةِ تكنولوجيا المعلوماتِ

Keywords:

: Driver drowsiness detection, Vision Transformer, recurrent neural networks, LSTM, GRU, BiLSTM, Advanced Driver Assistance Systems.

Abstract

Driver drowsiness during driving is considered one of the major factors contributing to traffic accidents, as it directly affects attention level, reaction time, and decision-making ability. This paper aims to develop an intelligent driver drowsiness detection system based on the analysis of facial frame sequences captured by an in-vehicle camera using deep learning techniques. The proposed system relies on the Vision Transformer (ViT) model to extract deep spatial features from facial images, which are then integrated with recurrent neural network models, namely Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional LSTM (BiLSTM), to model temporal dependencies and represent the evolution of the driver’s state over time. The system performance was evaluated using the public National Tsing Hua University - Driver Drowsiness Detection Dataset (NTHU-DDD), with the data divided into training, validation, and testing sets following a clear experimental protocol. Experimental results indicate that the ViT-LSTM model achieves the highest classification accuracy, while the ViT-GRU model demonstrates lower response time, making it more suitable for real-time applications. These findings confirm the effectiveness of the proposed architecture in achieving a balance between detection accuracy and temporal efficiency, highlighting its potential applicability within intelligent Advanced Driver Assistance Systems (ADAS).

Downloads

Published

2026-10-29