Research map / 3D vision / Face and shape
DiffComplete: Diffusion-based Generative 3D Shape Completion
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 |
|---|---|---|---|
| add occupancy masks to weight multi-scan feature fusion (vs simple average) (ablation) | DiffComplete · Two-condition fusion | l1 error: 0.13 -> 0.05 | Helped |
| add MLP psi finetuned for multi-condition fusion (ablation) | DiffComplete · Two-condition fusion | l1 error: 0.23 -> 0.05 | Helped |
| add a second partial scan as condition (ablation) | DiffComplete · 3D-EPN Chair multi-condition | l1 error: 0.07 -> 0.05 | Helped |
| add a third partial scan as condition (ablation) | DiffComplete · 3D-EPN Chair multi-condition | l1 error: 0.05 -> 0.04 | Helped |
| add more conditional scans (ablation) | DiffComplete · 3D-EPN Chair multi-condition | TMD: 0.025 -> 0.019 | Hurt |
| aggregate features at all four decoder levels instead of only 1/8 resolution (ablation) | DiffComplete · PatchComplete unseen | CD: 4.8 -> 4.1 | Helped |
| aggregate features at all four levels instead of only 1/8 (ablation) | DiffComplete · 3D-EPN Chair | MMD: 0.005 -> 0.002 | Helped |
| aggregate features at all four levels instead of only 1/8 (ablation) | DiffComplete · 3D-EPN Chair | TMD: 0.031 -> 0.025 | Hurt |
| use feature addition instead of cross-attention for aggregation (ablation) | DiffComplete · 3D-EPN Chair | l1 error: 0.12 -> 0.07 | Helped |
| use addition instead of concatenation (ablation) | DiffComplete · 3D-EPN Chair | l1 error: 0.07 -> 0.07 | No clear effect |
| DiffComplete hierarchical diffusion with control branch vs PatchComplete | PatchComplete · 3D-EPN known categories | l1 error: 0.088 -> 0.053 | Helped |
| DiffComplete on unseen real-scan categories vs PatchComplete | PatchComplete · PatchComplete real unseen | CD: 7.52 -> 7.18 | Helped |
Closest papers by meaning
- Diffusion 3D Features (Diff3F): Decorating Untextured Shapes with Distilled Semantic Features (CVPR 2024 · similarity 0.90)
- DinoComplete: 3D Shape Completion with Distilled Semantic Priors and State Space Models (NeurIPS 2026 · similarity 0.89)
- GeoComplete: Geometry-Aware Diffusion for Reference-Driven Image Completion (NeurIPS 2025 · similarity 0.88)
- PCDreamer: Point Cloud Completion Through Multi-view Diffusion Priors (CVPR 2025 · similarity 0.88)
- RaySt3R: Predicting Novel Depth Maps for Zero-Shot Object Completion (NeurIPS 2025 · similarity 0.87)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 2025 · 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.