GEOPARD: Geometric Pretraining for Articulation Prediction in 3D Shapes
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866917975840260096 |
|---|---|
| author | Goyal, Pradyumn Petrov, Dmitry Andrews, Sheldon Ben-Shabat, Yizhak Liu, Hsueh-Ti Derek Kalogerakis, Evangelos |
| author_facet | Goyal, Pradyumn Petrov, Dmitry Andrews, Sheldon Ben-Shabat, Yizhak Liu, Hsueh-Ti Derek Kalogerakis, Evangelos |
| contents | We present GEOPARD, a transformer-based architecture for predicting articulation from a single static snapshot of a 3D shape. The key idea of our method is a pretraining strategy that allows our transformer to learn plausible candidate articulations for 3D shapes based on a geometric-driven search without manual articulation annotation. The search automatically discovers physically valid part motions that do not cause detachments or collisions with other shape parts. Our experiments indicate that this geometric pretraining strategy, along with carefully designed choices in our transformer architecture, yields state-of-the-art results in articulation inference in the PartNet-Mobility dataset. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_02747 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | GEOPARD: Geometric Pretraining for Articulation Prediction in 3D Shapes Goyal, Pradyumn Petrov, Dmitry Andrews, Sheldon Ben-Shabat, Yizhak Liu, Hsueh-Ti Derek Kalogerakis, Evangelos Graphics We present GEOPARD, a transformer-based architecture for predicting articulation from a single static snapshot of a 3D shape. The key idea of our method is a pretraining strategy that allows our transformer to learn plausible candidate articulations for 3D shapes based on a geometric-driven search without manual articulation annotation. The search automatically discovers physically valid part motions that do not cause detachments or collisions with other shape parts. Our experiments indicate that this geometric pretraining strategy, along with carefully designed choices in our transformer architecture, yields state-of-the-art results in articulation inference in the PartNet-Mobility dataset. |
| title | GEOPARD: Geometric Pretraining for Articulation Prediction in 3D Shapes |
| topic | Graphics |
| url | https://arxiv.org/abs/2504.02747 |