Hyperbolic Part-Whole Image Segmentation

Mikhail Vlasenko, Mina Ghadimi Atigh, Pascal Mettes
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:2971-2979, 2026.

Abstract

Semantic segmentation typically focuses on pixel-level classification at the object level. Yet, objects naturally decompose into parts and subparts, mirroring human visual perception. In this work, we introduce a hyperbolic prototypical segmentation framework capable of simultaneously representing multiple granularity levels within a unified embedding space. Leveraging hyperbolic geometry’s unique capacity to model hierarchies effectively, we propose to embed class prototypes within the Poincar{é} ball. We introduce a tree-aware prototype initialization strategy and a distortion-\emph{p} loss that together yield improved hierarchical embeddings. Furthermore, we derive an optimized formulation of the hyperbolic distance function, enabling tractable inference for dense prediction tasks. A shared transformer encoder paired with separate hyperbolic heads allows efficient multi-level segmentation from a single model. Experiments on the recently introduced SubPartImageNet show that our approach (i) improves over the state-of-the-art, especially at the \emph{subpart} and \emph{part} levels, at a fraction of the number of parameters, (ii) enables zero-shot generalization, and (iii) allows for transfer from part- to object-level predictions without object-level supervision. All code is available at \url{https://github.com/mikhail-vlasenko/hyp-segmentation.}

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-vlasenko26a, title = { Hyperbolic Part-Whole Image Segmentation }, author = {Vlasenko, Mikhail and Atigh, Mina Ghadimi and Mettes, Pascal}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {2971--2979}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/vlasenko26a/vlasenko26a.pdf}, url = {https://proceedings.mlr.press/v300/vlasenko26a.html}, abstract = { Semantic segmentation typically focuses on pixel-level classification at the object level. Yet, objects naturally decompose into parts and subparts, mirroring human visual perception. In this work, we introduce a hyperbolic prototypical segmentation framework capable of simultaneously representing multiple granularity levels within a unified embedding space. Leveraging hyperbolic geometry’s unique capacity to model hierarchies effectively, we propose to embed class prototypes within the Poincar{é} ball. We introduce a tree-aware prototype initialization strategy and a distortion-\emph{p} loss that together yield improved hierarchical embeddings. Furthermore, we derive an optimized formulation of the hyperbolic distance function, enabling tractable inference for dense prediction tasks. A shared transformer encoder paired with separate hyperbolic heads allows efficient multi-level segmentation from a single model. Experiments on the recently introduced SubPartImageNet show that our approach (i) improves over the state-of-the-art, especially at the \emph{subpart} and \emph{part} levels, at a fraction of the number of parameters, (ii) enables zero-shot generalization, and (iii) allows for transfer from part- to object-level predictions without object-level supervision. All code is available at \url{https://github.com/mikhail-vlasenko/hyp-segmentation.} } }
Endnote
%0 Conference Paper %T Hyperbolic Part-Whole Image Segmentation %A Mikhail Vlasenko %A Mina Ghadimi Atigh %A Pascal Mettes %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-vlasenko26a %I PMLR %P 2971--2979 %U https://proceedings.mlr.press/v300/vlasenko26a.html %V 300 %X Semantic segmentation typically focuses on pixel-level classification at the object level. Yet, objects naturally decompose into parts and subparts, mirroring human visual perception. In this work, we introduce a hyperbolic prototypical segmentation framework capable of simultaneously representing multiple granularity levels within a unified embedding space. Leveraging hyperbolic geometry’s unique capacity to model hierarchies effectively, we propose to embed class prototypes within the Poincar{é} ball. We introduce a tree-aware prototype initialization strategy and a distortion-\emph{p} loss that together yield improved hierarchical embeddings. Furthermore, we derive an optimized formulation of the hyperbolic distance function, enabling tractable inference for dense prediction tasks. A shared transformer encoder paired with separate hyperbolic heads allows efficient multi-level segmentation from a single model. Experiments on the recently introduced SubPartImageNet show that our approach (i) improves over the state-of-the-art, especially at the \emph{subpart} and \emph{part} levels, at a fraction of the number of parameters, (ii) enables zero-shot generalization, and (iii) allows for transfer from part- to object-level predictions without object-level supervision. All code is available at \url{https://github.com/mikhail-vlasenko/hyp-segmentation.}
APA
Vlasenko, M., Atigh, M.G. & Mettes, P.. (2026). Hyperbolic Part-Whole Image Segmentation . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:2971-2979 Available from https://proceedings.mlr.press/v300/vlasenko26a.html.

Related Material