Deconvolutional Pixel Layer Model for Road segmentation without Human Assistance

The autonomous vehicle is the evolutionary goal of designing Advanced Driver Assistance System, (ADAS) to the point where human assistance is not needed anymore. To create a broad drivable geographical area and mapping for route planning we need robust and efficient semantic segmentation algorithm. Convolutional Neural Networks (CNN) have been able to achieve state of the art performance for tasks such as Image classification, Face recognition and Detection. However semantic segmentation has remained a challenging problem in the field of computer vision. With the help of convolution neural networks, we have witnessed prolific results over time. We propose a convolutional neural network model which uses Fully Convolutional Neural Network (FCN) with deconvolutional pixel layers. The goal is to create a hierarchy of features while the fully convolutional model does the primary learning and later deconvolutional model visually segments the target image. The proposed approach creates a direct link among the several adjacent pixels in the resulting feature maps. It also preserves the spatial features such as corners and edges in images and hence adding more accuracy to the resulting outputs. We test our algorithm on the Karlsruhe Institute of Technology and Toyota Technologies Institute (KITTI) street view datasets. Our method achieves a mIoU accuracy of 92.04 percent.