InpaintHuman: Reconstructing Occluded Humans with Multi-Scale UV Mapping and Identity-Preserving Diffusion Inpainting

2026-01-05

arXiv.org

AI Summary

Reconstructing complete and animatable 3D human avatars from monocular videos remains challenging under severe occlusions, as existing methods often suffer from geometric artifacts and temporal inconsistencies due to missing observations. This work proposes a novel approach that leverages multi-scale UV parameterization for robust geometry reconstruction and introduces an identity-preserving diffusion inpainting module, which integrates textual inversion and semantic guidance to effectively recover occluded regions while maintaining subject-specific details and temporal coherence. By employing hierarchical feature interpolation and pixel-level direct supervision, the method achieves significantly improved reconstruction quality on the PeopleSnapshot, ZJU-MoCap, and OcMotion datasets, yielding more complete geometries and temporally stable animations.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Humans and AI: Game Design — Virtual Humans, NPCs and Autonomous Characters
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Reconstructing complete and animatable 3D human avatars from monocular videos remains challenging, particularly under severe occlusions. While 3D Gaussian Splatting has enabled photorealistic human rendering, existing methods struggle with incomplete observations, often producing corrupted geometry and temporal inconsistencies. We present InpaintHuman, a novel method for generating high-fidelity, complete, and animatable avatars from occluded monocular videos. Our approach introduces two key innovations: (i) a multi-scale UV-parameterized representation with hierarchical coarse-to-fine feature interpolation, enabling robust reconstruction of occluded regions while preserving geometric details; and (ii) an identity-preserving diffusion inpainting module that integrates textual inversion with semantic-conditioned guidance for subject-specific, temporally coherent completion. Unlike SDS-based methods, our approach employs direct pixel-level supervision to ensure identity fidelity. Experiments on synthetic benchmarks (PeopleSnapshot, ZJU-MoCap) and real-world scenarios (OcMotion) demonstrate competitive performance with consistent improvements in reconstruction quality across diverse poses and viewpoints.

Problem

• 3D human reconstruction
• occlusion
• monocular video
• animatable avatars
• temporal consistency

Innovation

• multi-scale UV mapping
• identity-preserving diffusion inpainting
• 3D human reconstruction
• occlusion handling
• temporal coherence