Generative Semantic Scene Completion (GSSC) is a discrete-diffusion framework for outdoor LiDAR scene completion with three components: paired sparse-dense synthesis to generate training data, semantic-guided generative completion conditioned on sparse scans, and a refinement method called structured source discrete diffusion. The authors also contribute the PS3-SemanticKITTI dataset and reach state-of-the-art results on the SemanticKITTI leaderboard.
