Exploiting the Joint Potential of Instance Segmentation and Semantic Segmentation in Autonomous Driving

• Published in: 2023 International Conference for Advancement in Technology (ICONAT) pp. 1 - 7
• Main Authors: Usman, Mohammed, K, Thirumala Akash, Ahmed, Mohammed Riyaz, Gudodagi, Raveendra, T, Nitesh Kumar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 24.01.2023
• Subjects: Automobiles; Autonomous automobiles; Autonomous Vehicles; Deep learning; Deep Neural Networks; Instance segmentation; Neural networks; R-CNN; Satellites; Semantic segmentation; Transportation
• DOI: 10.1109/ICONAT57137.2023.10080167

The history of transportation spanning from 4000 BC to the present, is a mesmerizing subject that can deepen our understanding of travel from horses and camels, fixed-wheeled carts, riverboats, chariots, cycles, trains, airplanes, cars, and the development of the road locomotive. The future of transport is all about driverless cars or autonomous vehicles. The most crucial requirement for autonomous vehicles for navigation is to detect objects on the road. Autonomous vehicles provide potential answers for traffic congestion, road safety issues, and passenger comfort. The proposed model can detect and classify objects for assisting autonomous driving vehicles with the use of deep learning and neural network-based learning approaches. It aims to segment objects like people and automobiles through the Xception model, which carries out semantic segmentation, and Mask RCNN approaches, for instance, segmentation. Both these methods exhibit improvements in detection and are effective and relatively accurate with the results compared to Deep neural networks (DNN) in Contrast to Mask R-CNN and Xception model, respectively.