Research map / 3D vision / Face and shape
GEOPARD: Geometric Pretraining for Articulation Prediction in 3D Shapes
Pradyumn Goyal, Dmitry Petrov, Sheldon Andrews, Yizhak Ben-Shabat, Hsueh-Ti Derek Liu, Evangelos Kalogerakis
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 self-supervised pretraining on unannotated segmented shapes (ablation) | GEOPARD-u · PartNet-Mobility (unlabeled) | Axis Error AE (deg): 10.67 -> 9.18 | Improved |
| add geometric pretraining (ablation) | GEOPARD-u · PartNet-Mobility (unlabeled) | Revolute accuracy R-ACC: 0.87 -> 0.92 | Improved |
| add geometric pretraining (ablation) | GEOPARD-u · PartNet-Mobility (unlabeled) | Point Error PE: 0.06 -> 0.05 | Improved |
| aggregate part features with a learnable articulation query instead of mean pooling (ablation) | GEOPARD no pretraining · PartNet-Mobility (unlabeled) | Point Error PE: 0.11 -> 0.09 | Improved |
| regress per-point displacement votes for the pivot instead of predicting it from the articulation token (ablation) | GEOPARD no pretraining · PartNet-Mobility (unlabeled) | Point Error PE: 0.09 -> 0.06 | Improved |
| displacement voting for pivot (ablation) | GEOPARD no pretraining · PartNet-Mobility (unlabeled) | Axis Error AE (deg): 10.56 -> 10.67 | No clear effect |
| use GEOPARD instead of CAGE in labeled condition | GEOPARD · PartNet-Mobility (labeled) | Axis Error AE (deg): 11.41 -> 8.87 | Improved |
| use GEOPARD in labeled condition for motion type | GEOPARD · PartNet-Mobility (labeled) | Revolute accuracy R-ACC: 0.98 -> 0.98 | No clear effect |
| remove semantic part labels (unlabeled condition) | GEOPARD · PartNet-Mobility | Revolute accuracy R-ACC: 0.98 -> 0.92 | Hurt |
| order pretraining axes by PCA eigenvalue instead of motion range (ablation) | GEOPARD pretraining · PartNet-Mobility | training quality | Improved |
Closest papers by meaning
- Particulate: Feed-Forward 3D Object Articulation (CVPR 2026 · similarity 0.90)
- Articulation-Aware Canonical Surface Mapping (CVPR 2020 · similarity 0.87)
- Articulation in Motion: Prior-free Part Mobility Analysis for Articulated Objects By Dynamic-Static Disentanglement (ICLR 2026 · similarity 0.87)
- MagicArticulate: Make Your 3D Models Articulation-Ready (CVPR 2025 · similarity 0.87)
- ART: Articulated Reconstruction Transformer (CVPR 2026 · similarity 0.86)
- MeshArt: Generating Articulated Meshes with Structure-Guided Transformers (CVPR 2025 · similarity 0.86)
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.
