Closed liushawn618 closed 1 month ago
In two hand setting, the contact is not always in one hand. Your model needs to reason hand-object contact for the correct hand (see CDev in our paper). Also depth ambiguity (see MRRPE).
See our new paper: https://arxiv.org/abs/2403.16428
What difficulties might arise when transferring existing methods for single-hand and object joint reconstruction to joint reconstruction of two hands and an object?