Paper Title
REGRESSION-BASED DEEP LEARNING MODEL FOR ADAPTIVE DRIVING BEAM HEADLIGHTS

Abstract
The future of adaptive driving beam headlights (ADB), as the world moves toward automated driving (AD), is quickly coming into focus. To meet driver requirements for safety and visibility, engineers, developers, and designers are continuously researching the best combination of components. In automotive headlight systems, ADB automatically adjusts the beam pattern to provide the best visibility to the driver while reducing glare for oncoming drivers. Using cameras, sensors, and algorithms, the system detects the presence of other vehicles on the road and adjusts the headlight beams accordingly. As a result, the driver will have the highest level of visibility while minimizing the risk of dazzling other drivers. Many vehicles, including those in Europe, Asia and the Middle East, are equipped with ADB. In addition to revealing critical objects such as lane markings, pedestrians, and oncoming cars, adaptive capabilities help avoid temporarily blinding drivers of oncoming vehicles with full high beams. Designing and developing a solution for real road conditions is time-consuming, expensive, and complex. Thus, adaptive driving beam headlights are needed to detect oncoming vehicles and reduce glare for oncoming drivers. Fast, accurate, and easy-to-integrate detection solution is required for automotive vehicles. The purpose of this paper was to compare different detection methods that could be used to implement adaptive headlamps and to apply Machine Learning techniques for predicting fast and accurate object detection. CONCEPTS • Computing methodologies • Computer vision and pattern recognition • Object detection Additional Keywords and Phrases - Adaptive Driving Beam, Adaptive Front Lighting System, Convolutional Neural Network, Region-Based Deep Learning, Single Shot Multi Box Detector