DiffSSC: Semantic LiDAR Scan Completion using Denoising Diffusion Probabilistic Models

TL;DR

DiffSSC introduces a novel diffusion model approach for semantic LiDAR scan completion, effectively predicting unobserved scene geometry and semantics, leading to state-of-the-art results in autonomous driving datasets.

Contribution

This work extends diffusion models to semantic scene completion from LiDAR data, incorporating conditional generation and regularization techniques for improved accuracy.

Findings

Achieves state-of-the-art performance on autonomous driving datasets.

Outperforms existing methods in semantic scene completion.

Demonstrates effective diffusion-based completion of sparse LiDAR scans.

Abstract

Perception systems play a crucial role in autonomous driving, incorporating multiple sensors and corresponding computer vision algorithms. 3D LiDAR sensors are widely used to capture sparse point clouds of the vehicle's surroundings. However, such systems struggle to perceive occluded areas and gaps in the scene due to the sparsity of these point clouds and their lack of semantics. To address these challenges, Semantic Scene Completion (SSC) jointly predicts unobserved geometry and semantics in the scene given raw LiDAR measurements, aiming for a more complete scene representation. Building on promising results of diffusion models in image generation and super-resolution tasks, we propose their extension to SSC by implementing the noising and denoising diffusion processes in the point and semantic spaces individually. To control the generation, we employ semantic LiDAR point clouds as conditional input and design local and global regularization losses to stabilize the denoising process. We evaluate our approach on autonomous driving datasets, and it achieves state-of-the-art performance for SSC, surpassing most existing methods.