EUVIMEDEUVIMED entdecken
Crossref

RUMPL: Ray-Based Transformers for Universal Multi-View 2D to 3D Human Pose Lifting

Seyed Abolfazl Ghasemzadeh, Alexandre Alahi, Christophe De Vleeschouwer

IEEE Transactions on Image Processing · 2026 · 35 · 5509-5522

DOI: 10.1109/tip.2026.3694127 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Image Processing, 2026. Verfasst von Seyed Abolfazl Ghasemzadeh und 2 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Gefördert von: F?d?ration Wallonie-Bruxelles, Fonds De La Recherche Scientifique - FNRS.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Estimating 3D human poses from 2D images remains challenging due to occlusions and projective ambiguity. Multi-view learning-based approaches mitigate these issues but often fail to generalize to real-world scenarios, as large-scale multi-view datasets with 3D ground truth are scarce and captured under constrained conditions. To overcome this limitation, recent methods rely on 2D pose estimation combined with 2D-to-3D pose lifting trained on synthetic data. Building on our previous MPL framework, we propose RUMPL, a transformer-based 3D pose lifter that introduces a 3D ray-based representation of 2D keypoints. This formulation enables a model agnostic to camera parameters that can be universally deployed across arbitrary camera configurations in a given area without retraining or fine-tuning. A new View Fusion Transformer leverages learned fused-ray tokens to aggregate information along rays, further improving multi-view consistency. Evaluation on standard benchmarks shows that RUMPL significantly outperforms existing methods, yielding a 56.6% MPJPE (All KP) reduction on Human3.6M over triangulation-based methods and exceeding 70% improvement on the CMU Panoptic dataset when compared to transformer-based image-representation approaches. Results on new benchmarks, including in-the-wild multi-view and multi-person datasets, confirm its robustness and scalability.

Förderung

  • F?d?ration Wallonie-Bruxelles
  • Fonds De La Recherche Scientifique - FNRS

Zitieren

APA

Ghasemzadeh, S. A., Alahi, A., & De Vleeschouwer, C. (2026). RUMPL: Ray-Based Transformers for Universal Multi-View 2D to 3D Human Pose Lifting. IEEE Transactions on Image Processing, 35, 5509-5522. https://doi.org/10.1109/tip.2026.3694127

Vancouver

Ghasemzadeh SA, Alahi A, De Vleeschouwer C. RUMPL: Ray-Based Transformers for Universal Multi-View 2D to 3D Human Pose Lifting. IEEE Transactions on Image Processing. 2026;35:5509-5522. doi:10.1109/tip.2026.3694127

BibTeX
@article{Ghasemzadeh2026,
  author = {Ghasemzadeh, Seyed Abolfazl and Alahi, Alexandre and De Vleeschouwer, Christophe},
  title = {RUMPL: Ray-Based Transformers for Universal Multi-View 2D to 3D Human Pose Lifting},
  journal = {IEEE Transactions on Image Processing},
  year = {2026},
  volume = {35},
  pages = {5509--5522},
  doi = {10.1109/tip.2026.3694127}
}
Kostenlos starten →