Exploring Easy Boosts for Lidar Semantic Scene Completion
2026-06-02 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionRobotics
AI summaryⓘ
The authors explore easy ways to improve how lidar sensors understand and fill in 3D scenes without changing complex models. They show that adding simple labels created by existing tools to the lidar data helps current methods work better. They also improve results by marking which spaces are definitely empty or just unknown. With these simple tricks, older approaches can compete with or beat the newest ones in accuracy. The authors share their code for others to use.
lidarsemantic scene completionpoint cloudpseudo-labelsmIoUvisibility informationsemantic priorsoff-the-shelf segmentors3D scene understanding
Authors
Tetiana Martyniuk, Jonathan Seele, Alexandre Boulch, Gilles Puy, Renaud Marlet, Raoul de Charette
Abstract
This paper investigates "free lunch" strategies to boost the performance of lidar semantic scene completion (SSC) without requiring complex architectural redesigns. We first demonstrate that endowing input point clouds with semantic pseudo-labels from off-the-shelf segmentors significantly improves the performance of existing architectures. By evaluating these models against an oracle, we establish that high-quality semantic priors are a primary driver of mIoU gains. Furthermore, we equip the input lidar scan with visibility information that distinguishes between empty and unknown spaces, which provides a secondary performance boost across the tested architectures. Using these simple enhancements, we observe that older models remain competitive with state-of-the-art systems, and can even outperform them. Our code is available at https://github.com/astra-vision/SSC-Priors.