Research map / 3D vision / 3D generation
ScanComplete: Large-Scale Scene Completion and Semantic Segmentation for 3D Scans
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use ScanComplete completion-only instead of 3D-EPN | SUNCG scene completion · SUNCG test scenes | ℓ1 error (entire): 0.245 -> 0.182 | Helped |
| Jointly predict semantics with completion (ablation) | ScanComplete · SUNCG test scenes | ℓ1 error (entire): 0.182 -> 0.202 | Hurt |
| Add completion to semantic inference (ablation) | ScanComplete semantic labeling · SUNCG (vis) | average class accuracy: 43.6 -> 68.6 | Helped |
| Add the coarse-to-fine hierarchy (ablation) | ScanComplete semantic labeling · SUNCG (vis) | average class accuracy: 62.0 -> 68.6 | Helped |
| Add autoregressive voxel-group prediction (ablation) | Single-level deterministic completion · SUNCG test scenes | ℓ1 error (entire): 0.248 -> 0.204 | Helped |
| Train later hierarchy levels on predicted rather than ground-truth prior-level volumes (ablation) | Three-level ScanComplete · SUNCG test scenes | ℓ1 error (entire): 0.303 -> 0.182 | Helped |
| Use deterministic rather than probabilistic prediction (ablation) | Single-level autoregressive completion, input size 32 · SUNCG test scenes | ℓ1 error (entire): 0.218 -> 0.204 | Helped |
| Reduce probabilistic quantization bins from 256 to 32 (ablation) | Single-level autoregressive completion, input size 32 · SUNCG test scenes | ℓ1 error (entire): 0.218 -> 0.208 | Helped |
| Reduce probabilistic quantization bins from 32 to 8 (ablation) | Single-level autoregressive completion, input size 32 · SUNCG test scenes | ℓ1 error (entire): 0.208 -> 0.226 | Hurt |
| Fine-tune the semantic-only network on ScanNet (ablation) | ScanComplete semantic labeling · ScanNet test scenes | average class accuracy: 48.4 -> 58.7 | Helped |
| Use ScanComplete instead of the Matterport3D baseline | Matterport3D semantic labeling · Matterport3D test scenes | average class accuracy: 35.7 -> 48.3 | Helped |
Closest papers by meaning
- Context and Geometry Aware Voxel Transformer for Semantic Scene Completion (NeurIPS 2024 · similarity 0.89)
- DinoComplete: 3D Shape Completion with Distilled Semantic Priors and State Space Models (NeurIPS 2026 · similarity 0.88)
- Memory-Augmented Re-Completion for 3D Semantic Scene Completion (AAAI 2025 · similarity 0.88)
- Three Cars Approaching within 100m! Enhancing Distant Geometry by Tri-Axis Voxel Scanning for Camera-based Semantic Scene Completion (CVPR 2025 · similarity 0.87)
- Voxel Proposal Network via Multi-Frame Knowledge Distillation for Semantic Scene Completion (NeurIPS 2024 · similarity 0.87)
- H2GFormer: Horizontal-to-Global Voxel Transformer for 3D Semantic Scene Completion (AAAI 2024 · similarity 0.87)
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.