Research map / 3D vision / Face and shape
Disentangling Geometric Deformation Spaces in Generative Latent Shape Models
Tristan Aumentado‐Armstrong, Stavros Tsogkas, Sven Dickinson, Allan Jepson
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 shape-from-spectrum training (spectrum-derived intrinsics for reconstruction) (ablation) | GDVAE++ STD-S · SMAL | disentanglement score S: 0.49 -> 2.07 | Improved |
| add shape-from-spectrum training (ablation) | GDVAE++ FTL-S · SMAL | disentanglement score S: 0.43 -> 2.12 | Improved |
| add shape-from-spectrum training (ablation) | GDVAE++ STD-S · SMPL | disentanglement score S: 0.32 -> 2.58 | Improved |
| add shape-from-spectrum training (ablation) | GDVAE++ FTL-S · SMPL | disentanglement score S: 0.64 -> 2.46 | Improved |
| use mesh-derived LBO spectrum instead of point-cloud LBO (ablation) | GDVAE++ FTL-S · SMAL | disentanglement score S: 1.11 -> 2.12 | Improved |
| use mesh-derived LBO spectrum instead of point-cloud LBO (ablation) | GDVAE++ STD-S · SMPL | disentanglement score S: 1.43 -> 2.58 | Improved |
| add covariance and Jacobian disentanglement losses L_D (ablation) | GDVAE++ STD-S · SMAL | disentanglement score S: 2.14 -> 2.07 | No clear effect |
| add covariance and Jacobian disentanglement losses L_D (ablation) | GDVAE++ STD-S · SMPL | disentanglement score S: 2.01 -> 2.58 | Improved |
| add covariance and Jacobian disentanglement losses L_D (ablation) | GDVAE++ STD-S · SMAL | generative coverage: 0.47 -> 0.77 | Improved |
| add covariance and Jacobian disentanglement losses L_D (ablation) | GDVAE++ STD-S · SMAL | VAE reconstruction d_C: 0.44 -> 0.56 | Hurt |
| use FTL AE instead of STD AE (ablation) | autoencoder · SMAL | Chamfer reconstruction d_C: 0.35 -> 0.10 | Improved |
| use FTL AE instead of STD AE (ablation) | autoencoder · SMPL | latent rotation invariance C_X: 0.97 -> 0.71 | Hurt |
Closest papers by meaning
- Geometric Disentanglement for Generative Latent Shape Models (ICCV 2019 · similarity 0.94)
- Disentangled3D: Learning a 3D Generative Model With Disentangled Geometry and Appearance From Monocular Images (CVPR 2022 · similarity 0.90)
- Unsupervised Shape and Pose Disentanglement for 3D Meshes (ECCV 2020 · similarity 0.89)
- DISP6D: Disentangled Implicit Shape and Pose Learning for Scalable 6D Pose Estimation (ECCV 2022 · similarity 0.89)
- ShapeFlow: Learnable Deformation Flows Among 3D Shapes (NeurIPS 2020 · similarity 0.89)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 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.
