VolHuMe: a High-Resolution Large Scale Dataset of Volumetric Human Meshes

2026-06-22

AI Summary

Existing 4D human datasets suffer from insufficient geometric detail and texture resolution, limiting their utility for high-fidelity modeling research. This work presents a high-quality, large-scale 4D human scanning dataset captured using a close-range, high-resolution volumetric system comprising 64 RGB and 32 depth cameras. The dataset includes multi-view images, detailed meshes, skeletal rigs, point clouds, and semantic segmentations of clothing and facial regions for 104 subjects. By integrating SMPL-X fitting, high-resolution mesh reconstruction, and multimodal alignment techniques, the proposed pipeline substantially enhances local geometric fidelity and texture sharpness. As one of the most detailed 4D human datasets to date, it significantly outperforms existing benchmarks and exposes critical limitations in current human body reconstruction methods regarding fine-detail recovery.

Abstract

We introduce VolHuMe, a dataset of high-quality 4D human scans captured with a state-of-the-art volumetric studio using 64 RGB and 32 depth cameras. VolHuMe contains individual captures of 104 subjects and provides extensive ground truth, including SMPL-X, high-resolution meshes, multi-view RGB/depth images, rigged meshes, point clouds, garment segmentation, and detailed hand and facial geometry. Unlike prior datasets that primarily rely on full-body imagery, VolHuMe uses a close-range, high-resolution capture setup that preserves fine-grained body-part details, improving geometric fidelity and texture resolution. We benchmark VolHuMe on state-of-the-art methods across 3D and 4D human reconstruction tasks, showcasing the dataset's quality and exposing the limitations of current evaluation testbeds.

Problem

• 4D human reconstruction
• volumetric capture
• high-resolution meshes
• geometric fidelity
• human body modeling

Innovation

• volumetric capture
• high-resolution human meshes
• 4D human dataset
• multi-view RGB/depth imaging
• geometric fidelity