Research map / Point clouds / Segmentation and detection
Semantic Scene Completion Using Local Deep Implicit Functions on LiDAR Data
Christoph B. Rist, David Emmerichs, Markus Enzweiler, Dariu M. Gavrila
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add transposed-convolution upsampling in the encoder (vs CBN-only variant) (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | semantic mIoU: 23.8 -> 26.1 | Improved |
| add transposed-convolution upsampling in the encoder (vs CBN-only variant) (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | geometric completion IoU: 55.4 -> 57.8 | Improved |
| add multi-resolution feature maps in the decoder (vs highest resolution only) (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | semantic mIoU: 24.2 -> 26.1 | Improved |
| use CBN position conditioning instead of bilinear feature interpolation (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | semantic mIoU: 25.5 -> 26.1 | Improved |
| use baseline cell size 0.32 m instead of 150% cells (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | semantic mIoU: 24.1 -> 26.1 | Improved |
| use baseline cell size 0.32 m instead of 75% cells (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | geometric completion IoU: 54.1 -> 57.8 | Improved |
| use baseline cell size instead of 87.5% cells (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | semantic mIoU: 25.6 -> 26.1 | Improved |
| lower semantic loss weight, raise geometric weight (lambda_S=3.75, lambda_G=4) (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | geometric completion IoU: 57.8 -> 58.2 | Improved |
| lower semantic loss weight, raise geometric weight (lambda_S=3.75, lambda_G=4) (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | semantic mIoU: 26.1 -> 24.7 | Hurt |
| add consistency loss (lambda_C=1 vs 0) (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | semantic mIoU: 25.0 -> 26.1 | Improved |
| add semantic supervision (19 classes vs occupied/free only) (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | geometric completion IoU: 57.9 -> 57.8 | No clear effect |
| apply test-time augmentation (ablation) | Local-DIFs · SemanticKITTI validation (static scenes) | geometric completion IoU: 57.8 -> 58.5 | Improved |
Closest papers by meaning
- Towards Learning to Complete Anything in Lidar (ICML 2025 · similarity 0.92)
- SCPNet: Semantic Scene Completion on Point Cloud (CVPR 2023 · similarity 0.91)
- Semi-supervised 3D Semantic Scene Completion with 2D Vision Foundation Model Guidance (AAAI 2025 · similarity 0.89)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2025 · similarity 0.89)
- TALoS: Enhancing Semantic Scene Completion via Test-time Adaptation on the Line of Sight (NeurIPS 2024 · similarity 0.89)
- H2GFormer: Horizontal-to-Global Voxel Transformer for 3D Semantic Scene Completion (AAAI 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
