POEM: Reconstructing Hand in a Point Embedded Multi-view Stereo

Computer Vision and Pattern Recognition (CVPR), 2023

  Enable neural networks to capture 3D geometricalaware features is essential in multi-view based vision tasks. Previous methods usually encode the 3D information of multi-view stereo into the 2D features. In contrast, we present a novel method, named POEM, that directly operates on the 3D POints Embedded in the Multi-view stereo for reconstructing hand mesh in it. Point is a natural form of 3D information and an ideal medium for fusing features across views, as it has different projections on different views. Our method is thus in light ofa simple yet effective idea, that a complex 3D hand mesh can be represented by a set of 3D points that 1) are embedded in the multi-view stereo, 2) carry features from the multi-view images, and 3) wraps the hand in it. To leverage the power of points, we design two novel operations: point-based feature fusion and cross-set point attention mechanism. Evaluation on three challenging multi-view datasets shows that POEM outperforms the state-of-the-art in hand mesh reconstruction.


Recommended citation:

    author    = {Yang, Lixin and Xu, Jian and Zhong, Licheng and Zhan, Xinyu and Wang, Zhicheng and Wu, Kejian and Lu, Cewu},
    title     = {POEM: Reconstructing Hand in a Point Embedded Multi-View Stereo},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2023},
    pages     = {21108-21117}