SLCF-Net is a pioneering method that combines RGB images and sparse LiDAR scans to infer a 3D voxelized semantic scene. The model leverages Gaussian-decay Depth-prior Projection for feature projection and inter-frame feature propagation for temporal consistency. By integrating historical information, SLCF-Net excels in both accuracy and consistency metrics on the SemanticKITTI dataset. The model's performance surpasses other SSC baselines, showcasing its effectiveness in semantic scene completion tasks.
Till ett annat språk
från källinnehåll
arxiv.org
Djupare frågor