Enhancing Drivable Area Detection: A FCNN and VGG16 Approach for Autonomous Vehicles and ADAS

In recent days, multiple approaches are developed on autonomous vehicles as well as on Advance Driver Assistance System (ADAS) to minimize road accidents and to secure life. The drivable areas in many developing nations, as well as rural areas, are barely defined and maintained. Several existing approaches attempt to determine the drivable area by analyzing the lane mark and determining the drivable area. Traditional lane detection methods, on the other hand, can be difficult to conduct with a high detection rate in bad weather or in complex road situations. In this paper, we propose a novel method using Fully Convolutional Neural Network (FCNN) and Visual Geometry Group (VGG16) to detect drivable area. We describe a supervised learning strategy for simultaneously recognizing lane mark and drivable region. To begin, we applied a segmentation mask to the annotated image using FCNN and VGG16 to segment the drivable area. On KITTI datasets, we present extensive experimental results for drivable area detection