Research map / 3D vision / Face and shape
LAMP: Data-Efficient Linear Affine Weight-Space Models for Parameter-Controlled 3D Shape Generation and Extrapolation
Ghadi Nehme, Yanxia Zhang, Dule Shu, Matthew Klenk, Faez Ahmed
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 |
|---|---|---|---|
| train exemplar SDF decoders from a shared initialization (ablation) | LAMP · DrivAerNet++ single-parameter sweep, 100 exemplars | R^2: -37.77 -> 0.838 | Improved |
| train exemplar decoders with different random seeds (shared init kept) (ablation) | LAMP · DrivAerNet++ single-parameter sweep | R^2: 0.838 -> 0.773 | Hurt |
| increase number of exemplars from 10 to 100 (ablation) | LAMP · DrivAerNet++ | R^2: -7.289 -> 0.838 | Improved |
| increase number of exemplars from 100 to 1000 (ablation) | LAMP · DrivAerNet++ | R^2: 0.838 -> 0.862 | No clear effect |
| mix aligned SDF weights affinely instead of DNI weight prediction | LAMP · DrivAerNet++ large-range extrapolation, 4 parameters | R^2: -5.768 -> 0.867 | Improved |
| LAMP instead of AE-LPA with 1000 samples | LAMP with 100 samples · DrivAerNet++ interpolation | IoU (%): 92.63 -> 97.24 | Improved |
| LAMP instead of DNI for interpolation within dataset range | LAMP · DrivAerNet++ interpolation | IoU (%): 97.12 -> 97.24 | No clear effect |
| use a pretrained OpenShape embedding for the surrogate instead of random PointNet (ablation) | mesh-based surrogate · DrivAerNet++ parameter prediction | R^2 | Hurt |
| filter generations with linearity-mismatch threshold 0.01 | LAMP · human-annotated valid/invalid meshes | ROC AUC: ROC AUC 0.989, PR AUC 0.990 | Improved |
| add performance (drag) target to LAMP mixing | LAMP · DrivAerNet++ drag optimization | drag decay MAE (%): 5.2 (AE-LPA) -> 2.7 | Improved |
Closest papers by meaning
- 3DID: Direct 3D Inverse Design for Aerodynamics with Physics-Aware Optimization (NeurIPS 2025 · similarity 0.84)
- Flow-Corrected Shape Optimization: Taming Manifold Drift in High-Dimensional 3D Models (NeurIPS 2026 · similarity 0.84)
- ShapeShifter: 3D Variations Using Multiscale and Sparse Point-Voxel Diffusion (CVPR 2025 · similarity 0.83)
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · similarity 0.83)
- SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation (NeurIPS 2026 · similarity 0.83)
- CraftsMan3D: High-fidelity Mesh Generation with 3D Native Diffusion and Interactive Geometry Refiner (CVPR 2025 · similarity 0.83)
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.
