Login New user?  
Applied Mathematics & Information Sciences
An International Journal
               
 
 
 
 
 
 
 
 
 
 
 
 
 

Content
 

Volumes > Volume 17 > No. 5

 
   

Driver Mannersfinding Based On Transfer Learning And Deep Learning Networks

PP: 790-812
doi:10.18576/amis/170528
Author(s)
S Prema, K Radha, S Asokkumar,
Abstract
This paper proposes the driver behavior based on Transfer Learning (TL) and Deep Learning (DL).High concentration is required while driving, because it is possible that these behaviors are subordinated to other behaviors including eating, smoking, drinking, making phone calls, talking, adjusting the radio or being drowsy. Currently, these factors are responsible for the majority of traffic accidents. In order to avoid accidents, it is essential to develop applications that notify drivers in advance. In common classification networks, fully connected layers at the end are widely used. In this work the pre-processing technique is utilized to split the dataset and the classification module employs the suggested adaptive connections for extracting the feature maps. The classifier Convolutional Neural Network (CNN) module consists of a softmax layer and global average pool to determine the probability of each class. Here, the TL algorithm is adopted to improve performance of model and reduce time consumption. The overall architecture enhances the network variable and preserve a classification accuracy of 93.2%. The outcomes are implemented using Python platform. This proposed CNN based model worked well and successfully categorizes the driver's abnormal behaviors, according to analysis and results.

  Home   About us   News   Journals   Conferences Contact us