3D Multi-bodies: Fitting Sets of Plausible 3D Human Models to Ambiguous Image Data
This addresses the challenge of 3D human modeling in ambiguous scenarios for computer vision applications, representing an incremental improvement over existing methods.
The paper tackles the problem of obtaining dense 3D human reconstructions from ambiguous or occluded single views by recovering multiple plausible reconstructions, showing that their method outperforms alternatives on standard benchmarks and heavily occluded versions.
We consider the problem of obtaining dense 3D reconstructions of humans from single and partially occluded views. In such cases, the visual evidence is usually insufficient to identify a 3D reconstruction uniquely, so we aim at recovering several plausible reconstructions compatible with the input data. We suggest that ambiguities can be modelled more effectively by parametrizing the possible body shapes and poses via a suitable 3D model, such as SMPL for humans. We propose to learn a multi-hypothesis neural network regressor using a best-of-M loss, where each of the M hypotheses is constrained to lie on a manifold of plausible human poses by means of a generative model. We show that our method outperforms alternative approaches in ambiguous pose recovery on standard benchmarks for 3D humans, and in heavily occluded versions of these benchmarks.