English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We address the problem of fitting a parametric human body model (SMPL) topoint cloud data. Optimization-based methods require careful initialization andare prone to becoming trapped in local optima. Learning-based methods addressthis but do not generalize well when the input pose is far from those seenduring training. For rigid point clouds, remarkable generalization has beenachieved by leveraging SE(3)-equivariant networks, but these methods do notwork on articulated objects. In this work we extend this idea to human bodiesand propose ArtEq, a novel part-based SE(3)-equivariant neural architecture forSMPL model estimation from point clouds. Specifically, we learn a partdetection network by leveraging local SO(3) invariance, and regress shape andpose using articulated SE(3) shape-invariant and pose-equivariant networks, alltrained end-to-end. Our novel pose regression module leverages thepermutation-equivariant property of self-attention layers to preserverotational equivariance. Experimental results show that ArtEq generalizes toposes not seen during training, outperforming state-of-the-art methods by ~44%in terms of body reconstruction accuracy, without requiring an optimizationrefinement step. Furthermore, ArtEq is three orders of magnitude faster duringinference than prior work and has 97.3% fewer parameters. The code and modelare available for research purposes at https://arteq.is.tue.mpg.de.

Generalizing Neural Human Fitting to Unseen Poses With Articulated SE(3) Equivariance