MoMa: Skinned motion retargeting using masked pose modeling

Giulia Martinelli, Nicola Garau, Niccoló Bisagno, Nicola Conci

Abstract: Highlights•We introduce a skeleton-aware and shape-aware method for skinned motion retargeting.•We are first that can retarget between non-homeomorphic skeletons without paired motion.•We introduce a pose masking auto-encoder to reconstruct absent joints in the target.•We implement a novel a face-based optimizer to solve collisions in the body mesh.•We obtain state-of-the-art results on motion retargeting on the Mixamo dataset.