Research map / Point clouds / Shape and surface completion
LiteGE: Lightweight Geodesic Embedding for Efficient Geodesics Computation and Non-Isometric Shape Correspondence
Yohanes Yudhi Adikusuma, Qixing Huang, Ying He
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 |
|---|---|---|---|
| use LiteGE instead of NeuroGF on 300-point sparse clouds | geodesic regression · SMAL, 300 points | mean L1 error: 75 -> 3.4 | Improved |
| use LiteGE instead of NeuroGF on 2000-point clouds | geodesic regression · SMAL, 2000 points | mean L1 error: 13 -> 2.3 | Improved |
| use LiteGE instead of NeuroGF for batch queries | geodesic regression · 2K point clouds, N_Q=64 | memory (MB): 10344 -> 34.5 | Improved |
| use LiteGE instead of NeuroGF for batch queries | geodesic regression · 2K point clouds, N_Q=64 | inference time (ms): 1529 -> 13.4 | Improved |
| use LiteGE instead of NeuroGF on single query | geodesic regression · 2K point clouds, N_Q=1 | inference time (ms): 88 -> 1.9 | Improved |
| add 0.01-scale Gaussian noise to input point clouds | LiteGE · SMAL, 2K points | mean L1 error: 4.3 | No clear effect |
| use LiteGE instead of SMS on broken meshes | shape correspondence · Broken Mesh 5K | geodesic error: 28 -> 8 | Improved |
| use LiteGE instead of SMS on broken meshes | shape correspondence · Broken Mesh 5K | AUC: 33.2 -> 69.5 | Improved |
| use LiteGE instead of SMS on clean template meshes | shape correspondence · Template 4K | geodesic error: 2.2 -> 2.5 | Hurt |
| use LiteGE instead of SMS on remeshed 5K models | shape correspondence · Remeshed 5K | geodesic error: 7.44 -> 7.2 | Improved |
| use LiteGE instead of SMS on anisotropic 10K models | shape correspondence · Anisotropic 10K | geodesic error: 6.8 -> 7 | Hurt |
| increase point-cloud samples from 800 to 15,000 (ablation) | LiteGE path tracing · SMAL, 8 shapes | median prediction error: 4.73 -> 3.66 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- NeuroGF — NeuroGF: A Neural Representation for Fast Geodesic Distance and Path Queries (NeurIPS 2023) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- NeuroGF: A Neural Representation for Fast Geodesic Distance and Path Queries (NeurIPS 2023 · similarity 0.88)
- From Extrinsic to Intrinsic: Geodesic-Guided Representation Learning for 3D Geometric Data (ICML 2026 · similarity 0.87)
- GeoNet: Deep Geodesic Networks for Point Cloud Analysis (CVPR 2019 · similarity 0.87)
- Geodesic-HOF: 3D Reconstruction Without Cutting Corners (AAAI 2021 · similarity 0.86)
- Wormhole Loss for Partial Shape Matching (NeurIPS 2024 · similarity 0.85)
- APML: Adaptive Probabilistic Matching Loss for Robust 3D Point Cloud Reconstruction (NeurIPS 2025 · similarity 0.85)
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.
