DriveNet: A deep learning framework with attention mechanism for early driving maneuver prediction

Abstract

Inappropriate driving maneuvers are the leading cause of many car accidents. These accidents can be prevented if they are identified in advance and the driver is given the necessary assistance. Anticipating maneuvers is crucial for driving assistance systems in order to alert drivers and take appropriate measures to avoid or mitigate danger. In this paper, we introduce DriveNet a new approach that combines information about the driver’s behavior as well as the driving environment to predict the driving maneuvers. DriveNet utilizes a combination of convolutional neural network (CNN) and long short-term memory (LSTM) with attention mechanism to extract spatial information and capture long temporal dependencies. We evaluate DriveNet by performing a series of experiments using the publicly available Brain4Cars dataset. The findings show that the proposed approach achieves state-of-the-art performance and outperforms most previous methods. DriveNet has achieved an accuracy of 91.24%, a precision of 90.13%, and a recall of 91.44% for anticipation 4 seconds before the maneuvers occur.