Research map / 3D vision / Face and shape
Composite Shape Modeling via Latent Space Factorization
Anastasia Dubrovina, Fei Xia, Panos Achlioptas, Mira Shalah, Raphael Groscot, Leonidas J. Guibas
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 cycle consistency loss (ablation) | Decomposer-Composer · ShapeNet chairs (Mix) | Connectivity: 0.54 -> 0.65 | Improved |
| add cycle consistency loss (ablation) | Decomposer-Composer · ShapeNet chairs (Swap) | Connectivity: 0.62 -> 0.71 | Improved |
| add cycle consistency loss (ablation) | Decomposer-Composer · ShapeNet chairs (Rec) | Connectivity: 0.74 -> 0.82 | Improved |
| add cycle consistency loss (ablation) | Decomposer-Composer · ShapeNet chairs (Swap) | Classifier accuracy: 0.84 -> 0.89 | Improved |
| use learned projection matrices instead of fixed (ablation) | Decomposer-Composer · ShapeNet chairs (Mix) | Connectivity: 0.58 -> 0.65 | Improved |
| use learned projection matrices instead of fixed (ablation) | Decomposer-Composer · ShapeNet chairs (Mix) | Classifier accuracy: 0.77 -> 0.83 | Improved |
| add part decoder with STN to the composer (ablation) | Decomposer-Composer · ShapeNet chairs (Mix) | Connectivity: 0.23 -> 0.65 | Improved |
| add part decoder with STN to the composer (ablation) | Decomposer-Composer · ShapeNet chairs (Rec) | mIoU: 0.75 -> 0.64 | Hurt |
| add part decoder with STN to the composer (ablation) | Decomposer-Composer · ShapeNet chairs (Mix) | Classifier accuracy: 0.71 -> 0.83 | Improved |
| add cycle consistency loss (ablation) | Decomposer-Composer · ShapeNet chairs (Rec) | mIoU: 0.63 -> 0.64 | No clear effect |
| replace naive part placement with the full Decomposer-Composer | Naive placement · ShapeNet chairs (Mix) | Classifier accuracy: 0.21 -> 0.83 | Improved |
| replace ComplementMe with the full Decomposer-Composer | ComplementMe · ShapeNet chairs (Mix) | Connectivity: 0.47 -> 0.65 | Improved |
Closest papers by meaning
- StructEdit: Learning Structural Shape Variations (CVPR 2020 · similarity 0.89)
- Neural Parts: Learning Expressive 3D Shape Abstractions With Invertible Neural Networks (CVPR 2021 · similarity 0.88)
- ShapeFlow: Learnable Deformation Flows Among 3D Shapes (NeurIPS 2020 · similarity 0.88)
- Learning Part Generation and Assembly for Structure-Aware Shape Synthesis (AAAI 2020 · similarity 0.88)
- Neural Shape Deformation Priors (NeurIPS 2022 · similarity 0.88)
- Generative 3D Part Assembly via Part-Whole-Hierarchy Message Passing (CVPR 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.
