Research map / 3D vision / Face and shape
Latent Partition Implicit with Surface Codes for 3D Representation
Chao Chen, Yu-Shen Liu, Zhizhong Han
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 |
|---|---|---|---|
| blend surface codes with intrinsic (geodesic) distance affinity instead of uniform average (ablation) | LPI · D-FAUST (10 models) | L2CD x100: 0.0058 -> 0.0039 | Improved |
| blend with intrinsic distance instead of nearest-code one-hot affinity (ablation) | LPI · D-FAUST (10 models) | L2CD x100: 0.0060 -> 0.0039 | Improved |
| use intrinsic instead of Euclidean distance affinity (ablation) | LPI · D-FAUST (10 models) | L2CD x100: 0.0057 -> 0.0039 | Improved |
| use intrinsic instead of Euclidean distance affinity (ablation) | LPI · D-FAUST (100 models) | L2CD x100: 0.0064 -> 0.0059 | Improved |
| use intrinsic instead of Euclidean distance affinity (ablation) | LPI · D-FAUST (100 models) | L1CD x100: 0.698 -> 0.611 | Improved |
| use intrinsic instead of Euclidean distance affinity (ablation) | LPI · D-FAUST (10 models) | NC: 0.952 -> 0.952 | No clear effect |
| blend latent codes with Euclidean affinity instead of uniform average (ablation) | LPI · D-FAUST (10 models) | L2CD x100: 0.0058 -> 0.0057 | No clear effect |
| blend parts in latent space with surface codes versus voxel-grid latent codes (Nglod) | LPI · D-FAUST (100 models) | L2CD x100: 0.0062 -> 0.0059 | Improved |
| blend parts with d_E or d_G instead of nearest code only (ablation) | LPI · D-FAUST (10 models) | L2CD convergence | Improved |
| use LPI surface codes instead of NeuralPull network without codes | LPI · ShapeNet | L2CD x100: 0.038 -> 0.0171 | Improved |
| LPI versus NeuralParts for part reconstruction | LPI (parts) · D-FAUST (10 models) | IoU: 0.695 -> 0.828 | Improved |
| LPI versus NeuralParts for part reconstruction | LPI (parts) · D-FAUST (10 models) | NC: 0.906 -> 0.886 | Hurt |
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.
- NeuralParts — Neural Parts: Learning Expressive 3D Shape Abstractions With Invertible Neural Networks (CVPR 2021) · better on 2 of 2 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
- LP-DIF: Learning Local Pattern-Specific Deep Implicit Function for 3D Objects and Scenes (CVPR 2023 · similarity 0.90)
- Local Deep Implicit Functions for 3D Shape (CVPR 2020 · similarity 0.89)
- Discovering 3D Parts From Image Collections (ICCV 2021 · similarity 0.88)
- DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation (CVPR 2019 · similarity 0.88)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.87)
- LISR: Learning Linear 3D Implicit Surface Representation Using Compactly Supported Radial Basis Functions (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.
