Accurate road estimation is paramount for the successful operation of autonomous vehicles within their operational environment. Online mapping techniques play a crucial role in this pursuit by utilizing onboard sensors to estimate static objects such as lane markers, road edges, and pedestrian crossings in the vehicle's vicinity. Traditionally, models in this field have been trained using supervised learning methods. However, this project seeks to leverage recent advancements in semi/self-supervised learning techniques to enhance online mapping capabilities. To accomplish this, not only labeled datasets but also a substantial amount of unlabeled data are required. Consequently, there is a need for additional storage resources to accommodate the vast amount of data necessary for training and improving the mapping algorithms.