Driver Activity Recognition for Vehicles Using Deep CNN Models

  • R Gnanavel, G Kalpana, MR Ravisankar, K Naveen Kumar, NS Ramasamy

Abstract

Most Important factors that affect car passenger safety are Driver decisions and behaviours. Common activities that the drivers are involved in include – Safe driving[1], texting on mobile phones[2], speaking on mobile phone[3], operating the radio[4], drinking or eating[5], reaching behind[6], adjusting makeup or hair[7], or talking to his/her fellow passengers[8].  Among these activities, the first four are regarded as normal driving tasks, while the last seven are classified into the distracted group. The idea of the approach is to receive the live video feed from the dashboard camera, analyze the frames using pre-trained CNN models then fine tune it using transfer learning method, to determine the type of activity that the driver is being involved in. The objective of the project is to observe the driver behaviours, using which a driver activity recognition system was built using deep Convolutional Neural Networks (CNN).

Published
2020-05-28
How to Cite
R Gnanavel, G Kalpana, MR Ravisankar, K Naveen Kumar, NS Ramasamy. (2020). Driver Activity Recognition for Vehicles Using Deep CNN Models. International Journal of Control and Automation, 13(4), 745 - 759. Retrieved from https://sersc.org/journals/index.php/IJCA/article/view/18868
Section
Articles