Computer scientists developed a deep learning method to create realistic objects for virtual environments that can be used to train robots. The researchers used TACC’s Maverick2 supercomputer to train the generative adversarial network. The network is the first that can produce colored point clouds with fine details at multiple resolutions.
