Summary
Dense4D, introduced by Rıza Alp Güler (arXiv 2610.22325), is the first feed-forward model dedicated to dense 4D correspondence across the human body, clothing, and hair from a set of monocular RGB images, without requiring a body template. The model jointly processes person-centred crops with known image-to-crop transforms to predict Human Point Maps (HPMs) and per-image depth. An HPM assigns each person pixel the 3D position of its observed surface point at a common reference time, enabling amodal correspondence: surface points hidden in the reference image can still be localized from observations in other frames. This addresses the core challenge of reconstructing human motion in 4D—identifying the same surface points through articulation, deformation, and visibility changes. Experiments on real human subjects show that Dense4D substantially outperforms evaluated baselines in tracking accuracy and consistency with scanned surfaces. Project page: https://alpguler.com/dense4d
Paper Overview
Field: Computer Vision
Author: Rıza Alp Güler
Published: 2026-10-15
arXiv: 2610.22325
Abstract (Translated)
Reconstructing human motion in 4D requires identifying the same surface points through articulation, deformation and changes in visibility. We introduce Dense4D, to our knowledge the first feed-forward model dedicated to dense 4D correspondence across the body, clothing and hair from a set of monocular RGB images, without a body template.
Trained specifically for human geometry, Dense4D jointly processes person-centred crops with known image-to-crop transforms to predict Human Point Maps (HPMs) and depth for each image. An HPM assigns each person pixel the 3D position of its observed surface point at a common reference time. These maps support amodal correspondence: points hidden in the reference image can be localised from observations in other frames. Matching HPM coordinates across crops yields dense, template-free correspondence over time.
Experiments on real human subjects demonstrate that Dense4D substantially outperforms evaluated baselines in both tracking accuracy and consistency with scanned surfaces.
Project page: https://alpguler.com/dense4d
*Auto-collected on 2026-10-17.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178636102