Connected and Autonomous Vehicles (CAVs) are vehicles that provide connectivity between other vehicles (V2V), infrastructure (V2I) and any things (V2X) using various communication technologies. Deploying CA V s can make transportation safer, improve mobility and provide benefits to the Smart city environment. For autonomous driving, lane detection/segmentation is one of important tasks, and changing lanes is one of the crucial driving decisions. This paper exclusively investigates drivable space segmentation and state-of-the-art deep learning model for instance segmentation. The results show that the selected Mask R-CNN model accurately detects and segments direct lane and alternative lanes with high confidence score.