Deep Learning-Based Driver Activity Recognition Using Diverse Driver Profiles
Driver attentiveness is paramount for ensuring both driver’s personal safety and for road safety. Annually, numerous lives are tragically lost due to driver distraction, underscoring the necessity for a driver activity monitoring system to enable prompt action and response. This research...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10740280/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Driver attentiveness is paramount for ensuring both driver’s personal safety and for road safety. Annually, numerous lives are tragically lost due to driver distraction, underscoring the necessity for a driver activity monitoring system to enable prompt action and response. This research study presents a framework that combines computer vision with deep learning for driver activity recognition. This study makes two novel contributions: a dataset having wide range of driver activity classes and implementation of two deep learning models- 2D Convolutional Long Short-Term Memory (2D ConvLSTM) network and a Long Recurrent Convolutional Network (LRCN)- specifically designed for driver activity recognition. The dataset is collected in a naturalistic environment, featuring a diverse group of drivers with balanced representation of both genders and specifically Hijabi and veiled drivers that are underrepresented in existing datasets. The experiments in our study were conducted using a systematic approach aimed at analyzing and improving the generalizability of the constructed models when exposed to diverse groups of individuals. Our models achieved an accuracy of 97.99% for activities recognition of drivers. Employing the constructed models can substantially contribute to designing modern Advanced Driver Assistance Systems that to-date possess certain biases which our framework evidently addresses. |
|---|---|
| ISSN: | 2169-3536 |