Research map / Point clouds / Shape and surface completion
Learning Geometry-Aware Embedding Fields for Intrinsic Riemannian Mappings
Bo Pang, Simone Foti, Tolga Birdal
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 geometric-prior loss terms beyond the supervised loss (ablation) | Embedding-field log map · ShapeNet ablation (100 models) | EMDD: 0.162 -> 0.112 | Improved |
| add geometric-prior loss terms beyond the supervised loss (ablation) | Embedding-field log map · ShapeNet ablation (100 models) | angular error (deg): 11.5 -> 11.2 | No clear effect |
| add geometric-prior loss terms beyond the supervised loss (ablation) | Embedding-field exp map · ShapeNet ablation (100 models) | Exp error: 0.025 -> 0.028 | Hurt |
| add training data filtering (ablation) | Embedding-field model · ShapeNet ablation (100 models) | relative training cost: 1.42x -> 1.00x | Improved |
| add training data filtering (ablation) | Embedding-field model · ShapeNet ablation (100 models) | EMDD: 0.112 -> 0.112 | No clear effect |
| replace Heat method with learned embedding field | Log map · ShapeNet | angular error (deg): 16.7 -> 11.1 | Improved |
| replace Heat method with learned embedding field | Log map · Noisy ShapeNet | EMDD: 0.326 -> 0.202 | Improved |
| replace projection baseline with learned embedding field | Log map · Raw ShapeNet | EMDD: 0.166 -> 0.169 | No clear effect |
| replace geodesic tracing with learned embedding field | Exp map · ShapeNet | Exp error: 0.036 -> 0.026 | Improved |
| replace geodesic tracing with learned embedding field | Exp map · DFAUST | Exp error: 0.033 -> 0.043 | Hurt |
| replace projection baseline with learned embedding field | Exp map · Kinect-v2 | Exp error: 0.064 -> 0.060 | Mixed |
| replace Heat method with learned embedding field | Log map · DFAUST | angular error (deg): 25.47 -> 16.51 | Improved |
Closest papers by meaning
- From Extrinsic to Intrinsic: Geodesic-Guided Representation Learning for 3D Geometric Data (ICML 2026 · similarity 0.87)
- Robust Hyperbolic Learning with Curvature-Aware Optimization (NeurIPS 2025 · similarity 0.85)
- Grounding Continuous Representations in Geometry: Equivariant Neural Fields (ICLR 2025 · similarity 0.85)
- Follow the Energy, Find the Path: Riemannian Metrics from Energy-Based Models (NeurIPS 2025 · similarity 0.84)
- Neural Isometries: Taming Transformations for Equivariant ML (NeurIPS 2024 · similarity 0.84)
- Euclidean Embedding of Data Using Local Distances (NeurIPS 2026 · similarity 0.84)
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.
