# InstantHMR开源移动端人体网格恢复：S23 NPU推理仅5.5ms

> 原文链接：https://robotics.hellobike.com/papers/2026-11-06-2611-09195-d14a8d45
> 频道：papers | 更新日期：2026-11-06 | 标签：auto, arxiv, hf-daily-papers
> 核心摘要：面向机器人与AR/VR的开源单目人体网格回归模型InstantHMR发布，基于MHR参数化模型与SAM 3D Body数据训练。在三星S23 NPU上实现6890顶点网格5.5ms实时推理，精度达40.0mm PA-MPJPE，速度较现有移动端方案提升12倍。

# InstantHMR开源移动端人体网格恢复：S23 NPU推理仅5.5ms

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[InstantHMR : Real-Time Human Mesh Recovery on Your Phone](https://arxiv.org/abs/2611.09195)（HF Daily Papers · arXiv 2611.09195）

## 核心内容

针对移动端实时3D人体网格恢复难题，研究团队推出开源模型InstantHMR，专为机器人、生物力学及AR/VR等具身交互场景设计。该模型基于Momentum Human Rig (MHR) 参数化模型构建，仅使用SAM 3D Body公开数据集训练。在三星Galaxy S23 NPU上实测，单人6890顶点身体网格恢复耗时仅5.5ms，NVIDIA RTX 4070 GPU端低至0.98ms，刷新了现有方法的推理速度记录。

InstantHMR在保持极速推理的同时兼顾精度，在3DPW、EMDB和RICH基准上的PA-MPJPE分别为40.0mm、48.8mm和41.7mm，与高精度模型持平。相比唯一兼容移动端且精度略优的竞品，其速度快12倍。团队还建立了统一评测协议，复现并对比了包括Meta官方MHR-to-SMPL转换在内的11种SOTA方法，覆盖桌面TensorRT及两款手机NPU共八个硬件平台，代码与模型已在GitHub开源。

## 关键看点

- **极致端侧推理**：三星S23 NPU单帧5.5ms恢复6890顶点网格，比现有移动端兼容方案快12倍，满足机器人实时感知需求
- **高精度保持**：3DPW基准PA-MPJPE达40.0mm，EMDB为48.8mm，在毫秒级推理下仍维持与SOTA相当的几何重建精度
- **统一评测协议**：重新评估11种SOTA方法并包含Meta官方转换工具，横跨桌面GPU与手机NPU等8个平台提供可复现基准
- **全栈开源资产**：基于MHR模型与SAM 3D Body数据训练，代码与预训练权重已在GitHub开放，降低具身视觉算法部署门槛

## 工程与产业观察

InstantHMR将高精度人体网格恢复压缩至移动端NPU 5.5ms延迟，解决了具身智能人机交互中实时姿态感知的算力瓶颈。其统一评测协议与开源权重为机器人开发者提供了标准化的端侧视觉基线，使低成本移动设备也能胜任高动态场景下的精准人体建模任务。

## 英文摘要 (Abstract)

Real-time 3D human mesh recovery from a single image on mobile devices is a challenging problem with many applications, including robotics, biomechanics, AR/VR, and exergames. Current state-of-the-art models are either too slow to run efficiently on mobile devices or sacrifice accuracy for speed. We present InstantHMR, a family of open-source single-image regressors suited for mobile inference, based on the Momentum Human Rig (MHR) parametric model and trained exclusively on publicly available annotations from the SAM 3D Body dataset. Our approach achieves remarkable performance on a Samsung Galaxy S23 NPU, recovering a 6,890-vertex body mesh in 5.5 ms per person, and 0.98 ms on an NVIDIA RTX 4070 GPU. This speed surpasses all existing methods we evaluated. On mobile, it is 12x faster than the single mobile-compatible method that slightly outperforms ours in accuracy, while achieving competitive accuracy with 40.0 mm PA-MPJPE on 3DPW, 48.8 mm on EMDB and 41.7 mm on RICH. To ensure fair comparisons, we re-evaluate 11 state-of-the-art methods under a unified protocol aligned with the closest MHR-based competitors, including Meta's official MHR-to-SMPL conversion, and report their reproduction results. Additionally, we measure latency across eight different targets, ranging from desktop TensorRT to two phone NPUs. Code and models are available at https://github.com/Naturalpad-studio/InstantHMR.

Hugging Face Daily Papers 热度：2 upvotes
作者：Mohamed Adjel, Kahina Chalabi, Vincent Bonnet
:::

## 原始参考来源 (Sources)

- [HF Daily Papers · arXiv 2611.09195：InstantHMR : Real-Time Human Mesh Recovery on Your Phone](https://arxiv.org/abs/2611.09195)