1 Introduction

Introduction to Driver Behavior recognition.

When considering dangerous situations in traffic we cannot only focus on dectecting event such as collission or near misses. It is very important to be able to detect different kind of behaviors that could lead to these events, most of them taking place within the vehicles and related to drivers behavior.

Taking this into account, our application consists in showcasting the advantages of the detection models within Intel's OpenVINO toolkit applied for recognizing driver's behavior looking to be able to anticipate potentially incidents due to, i.e, drowsiness status, scattered attention, etc.

In this sense, we also use external detection models that might be optimized when being integrated with OpenVINO, that will help us to executing the following scenarios:

1. Drowsiness State Detection

• Yawn detection
• Blink counting

2. Distraction State Detection

• Head Position detection
• Gaze Attention detection

Being our main purpose to develop an application that allows taking preventive actions over dangerous situations in traffic, we're aiming to integrate this detection models with a notification/alarm system that will alert the driver and will help him/her to recover and drive safetly. According to this, in first instance, we want to explore the advantages of this app aplying it to fleet management, as part of our effort on looking foward a success use case.