Etudiant : SASSIOUI ABDELLATIF

Filière : Master Big Data Analytics & Smart Systems (BDSaS)

Encadrant : Pr. BOUHOUTE AFAF

Annèe : 2022

Résumé : Inappropriate driving maneuvers are the main cause behind many car accidents. These accidents can be avoided if they are detected beforehand and the driver is assisted ac- cordingly by Advanced Driver Assistant System ADAS. Recent studies have focused on predicting driver intention as a key part of these systems. Anticipating manoeuvres in advance alerts drivers and gives drivers more time to avoid or prepare for danger. To overcome this challenge, we propose in this work a system that anticipates maneuvers by predicting the driver’s intention using a dataset which consists of driver videos in differ- ent situations and maneuvers, as well as information on the environment, such as speed, empty lines, and the existence of an artifact. We introduce a novel method that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) with At- tention Mechanism to extract spatial features and capture long temporal dependencies. We perform experiments on the publicly available Brain4Cars dataset.The findings show that our proposed method outperforms the majority of previous methods and achieves state-of-the-art performance. Our approach can anticipate maneuvers 4 seconds before they occur with the accuracy of 91.24%, precision of 90.13% and recall of 91.44%