We introduce ExPose, a novel extreme pose estimation framework via off-the-shelf 3D foundation models with an aid of pose-rewarded reinforcement learning of video generation models.

BibTeX

@InProceedings{Yoon_2026_CVPR,
    author    = {Yoon, Youngho and Cho, Wonjune and Ha, Hyunho and Kim, Sujung and Yoon, Kuk-Jin},
    title     = {ExPose: Reinforcing Video Generation Models for Extreme Pose Estimation},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2026},
    pages     = {32636-32646}
}