Pith. sign in

REVIEW

SeSDF: Self-evolved Signed Distance Field for Implicit 3D Clothed Human Reconstruction

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2304.00359 v1 pith:BSM2WSUT submitted 2023-04-01 cs.CV

SeSDF: Self-evolved Signed Distance Field for Implicit 3D Clothed Human Reconstruction

classification cs.CV
keywords humanclothedframeworkmodelreconstructiondistancefieldimages
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We address the problem of clothed human reconstruction from a single image or uncalibrated multi-view images. Existing methods struggle with reconstructing detailed geometry of a clothed human and often require a calibrated setting for multi-view reconstruction. We propose a flexible framework which, by leveraging the parametric SMPL-X model, can take an arbitrary number of input images to reconstruct a clothed human model under an uncalibrated setting. At the core of our framework is our novel self-evolved signed distance field (SeSDF) module which allows the framework to learn to deform the signed distance field (SDF) derived from the fitted SMPL-X model, such that detailed geometry reflecting the actual clothed human can be encoded for better reconstruction. Besides, we propose a simple method for self-calibration of multi-view images via the fitted SMPL-X parameters. This lifts the requirement of tedious manual calibration and largely increases the flexibility of our method. Further, we introduce an effective occlusion-aware feature fusion strategy to account for the most useful features to reconstruct the human model. We thoroughly evaluate our framework on public benchmarks, demonstrating significant superiority over the state-of-the-arts both qualitatively and quantitatively.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.