Human4K: A Large-Scale 4K Multi-View Mocap Dataset for Whole-Body 3D Human Reconstruction

2026-07-15

AI Summary

Existing 3D human reconstruction methods often underperform in complex real-world scenarios due to geometric instability, inaccurate joint estimation, and self-occlusion, primarily caused by the lack of high-resolution, high-fidelity training data with diverse poses. To address this gap, this work introduces Human4K, a novel dataset that uniquely combines eight-view synchronized 4K video, professional Vicon motion capture, and highly self-occluded full-body motions across 11 subjects, yielding over six million frames. High-fidelity SMPL-X annotations are generated via a custom Motion Retargeting and Refinement Module (MRRM). Models trained on Human4K achieve significantly improved reconstruction accuracy on standard benchmarks, with notable gains in challenging regions such as hands, feet, and depth-ambiguous areas, thereby filling a critical void in high-fidelity full-body 3D human reconstruction data.

Abstract

Recent advances in 3D human reconstruction have improved overall performance, yet current models still fail in the most challenging real-world scenarios. They often produce unstable geometry, inaccurate limb articulation and unreliable predictions under depth ambiguity or self-occlusion. A key reason is that existing datasets still lack the combination of high-resolution images, high-precision annotations and diverse whole-body motions required to support robust reconstruction. To address this gap, we present Human4K, a large-scale 4K multi-view whole-body human reconstruction dataset with mocap-accurate SMPL-X annotations. Human4K contains over six million 4K images captured by an eight-view high-resolution camera system synchronized with a professional Vicon motion capture setup, covering 11 subjects performing complex, highly articulated and strongly self-occluded full-body motions. All sequences are processed by a Motion-Retargeting and Refinement Module (MRRM) to ensure precise alignment for the full body and extremities. Experimental results show that training with Human4K consistently improves whole-body reconstruction on standard benchmarks, with particularly large gains for hands, feet and depth-ambiguous limb configurations.

Problem

• 3D human reconstruction
• self-occlusion
• depth ambiguity
• high-resolution dataset
• whole-body motion

Innovation

• 4K multi-view dataset
• SMPL-X annotations
• motion capture
• self-occlusion handling
• whole-body reconstruction