Outdoors displacement is among the navigation tasks of mobile autonomous devices.Such displacement may take place in particular environments, like rural roads, characterized by a restricted set of scene elements that surround the displacements. Road scene segmentation is important for autonomous driving and pedestrian detection. This project proposes to implement a road segmentation algorithm using a Fully Convolutional Network (FCN). In this project, FCN-VGG16 is implemented and trained with KITTI dataset for road segmentation. Then we compare it to the well known deeplab algorithm for segmentation trained on cityscapes dataset.
Code and results can be found in google colaboratory notebook