MetaCap: Meta-learning Priors from Multi-View Imagery for Sparse-view Human Performance Capture and Rendering

arXiv:2403.18820v1 Announce Type: new
Abstract: Faithful human performance capture and free-view rendering from sparse RGB observations is a long-standing problem in Vision and Graphics. The main challenges are the lack of observations and the inherent ambiguities of the setting, e.g. occlusions and depth ambiguity. As a result, radiance fields, which have shown great promise in capturing high-frequency appearance and geometry details in dense setups, perform poorly when na\"ively supervising them on sparse camera views, as the field simply overfits to …

abstract arxiv challenges cs.cv free graphics human human performance meta meta-learning performance rendering type view vision