METHOD AND APPARATUS FOR SEMANTIC SEGMENTATION AND DEPTH COMPLETION USING A CONVOLUTIONAL NEURAL NETWORK

摘要

A computer implemented method for generating semantically segmented images and depth completion images using a convolutional neural network (CNN) from input visual images and/or input depth images. The central elements of a CNN for semantic segmentation and depth completion are: (i) given both an input visual image and an input depth image, (ii) given only an input visual image, or (iii) given only an input depth image. Given any combinations of these input images, there is a common expression that allows both tasks to be performed.