ScrewSplat: An End-to-End Method for Articulated Object Recognition

Seungyeon Kim, Junsu HA, Young Hun Kim, Yonghyeon Lee, Frank C. Park
Proceedings of The 9th Conference on Robot Learning, PMLR 305:309-335, 2025.

Abstract

Articulated object recognition – the task of identifying both the geometry and kinematic joints of objects with movable parts – is essential for enabling robots to interact with everyday objects such as doors and laptops. However, existing approaches often rely on strong assumptions, such as a known number of articulated parts; require additional inputs, such as depth images; or involve complex intermediate steps that can introduce potential errors – limiting their practicality in real-world settings. In this paper, we introduce **ScrewSplat**, a simple end-to-end method that operates solely on RGB observations. Our approach begins by randomly initializing screw axes, which are then iteratively optimized to recover the object’s underlying kinematic structure. By integrating with Gaussian Splatting, we simultaneously reconstruct the 3D geometry and segment the object into rigid, movable parts. We demonstrate that our method achieves state-of-the-art recognition accuracy across a diverse set of articulated objects, and further enables zero-shot, text-guided manipulation using the recovered kinematic model.

Cite this Paper

BibTeX
@InProceedings{pmlr-v305-kim25a, title = {ScrewSplat: An End-to-End Method for Articulated Object Recognition}, author = {Kim, Seungyeon and HA, Junsu and Kim, Young Hun and Lee, Yonghyeon and Park, Frank C.}, booktitle = {Proceedings of The 9th Conference on Robot Learning}, pages = {309--335}, year = {2025}, editor = {Lim, Joseph and Song, Shuran and Park, Hae-Won}, volume = {305}, series = {Proceedings of Machine Learning Research}, month = {27--30 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v305/main/assets/kim25a/kim25a.pdf}, url = {https://proceedings.mlr.press/v305/kim25a.html}, abstract = {Articulated object recognition – the task of identifying both the geometry and kinematic joints of objects with movable parts – is essential for enabling robots to interact with everyday objects such as doors and laptops. However, existing approaches often rely on strong assumptions, such as a known number of articulated parts; require additional inputs, such as depth images; or involve complex intermediate steps that can introduce potential errors – limiting their practicality in real-world settings. In this paper, we introduce **ScrewSplat**, a simple end-to-end method that operates solely on RGB observations. Our approach begins by randomly initializing screw axes, which are then iteratively optimized to recover the object’s underlying kinematic structure. By integrating with Gaussian Splatting, we simultaneously reconstruct the 3D geometry and segment the object into rigid, movable parts. We demonstrate that our method achieves state-of-the-art recognition accuracy across a diverse set of articulated objects, and further enables zero-shot, text-guided manipulation using the recovered kinematic model.} }
Endnote
%0 Conference Paper %T ScrewSplat: An End-to-End Method for Articulated Object Recognition %A Seungyeon Kim %A Junsu HA %A Young Hun Kim %A Yonghyeon Lee %A Frank C. Park %B Proceedings of The 9th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Joseph Lim %E Shuran Song %E Hae-Won Park %F pmlr-v305-kim25a %I PMLR %P 309--335 %U https://proceedings.mlr.press/v305/kim25a.html %V 305 %X Articulated object recognition – the task of identifying both the geometry and kinematic joints of objects with movable parts – is essential for enabling robots to interact with everyday objects such as doors and laptops. However, existing approaches often rely on strong assumptions, such as a known number of articulated parts; require additional inputs, such as depth images; or involve complex intermediate steps that can introduce potential errors – limiting their practicality in real-world settings. In this paper, we introduce **ScrewSplat**, a simple end-to-end method that operates solely on RGB observations. Our approach begins by randomly initializing screw axes, which are then iteratively optimized to recover the object’s underlying kinematic structure. By integrating with Gaussian Splatting, we simultaneously reconstruct the 3D geometry and segment the object into rigid, movable parts. We demonstrate that our method achieves state-of-the-art recognition accuracy across a diverse set of articulated objects, and further enables zero-shot, text-guided manipulation using the recovered kinematic model.
APA
Kim, S., HA, J., Kim, Y.H., Lee, Y. & Park, F.C.. (2025). ScrewSplat: An End-to-End Method for Articulated Object Recognition. Proceedings of The 9th Conference on Robot Learning, in Proceedings of Machine Learning Research 305:309-335 Available from https://proceedings.mlr.press/v305/kim25a.html.

Related Material