Mem-MLP: Real-Time 3D Human Motion Generation from Sparse Inputs

TL;DR

Mem-MLP introduces a real-time neural network model with a novel Memory-Block for accurate 3D human motion generation from sparse sensor data, enhancing AR/VR applications with improved consistency and speed.

Contribution

The paper presents a new MLP-based approach with Memory-Block and multi-task learning for efficient, accurate 3D motion reconstruction from limited inputs.

Findings

Outperforms state-of-the-art baselines in accuracy

Achieves 72 FPS on mobile HMDs

Reduces prediction errors significantly

Abstract

Realistic and smooth full-body tracking is crucial for immersive AR/VR applications. Existing systems primarily track head and hands via Head Mounted Devices (HMDs) and controllers, making the 3D full-body reconstruction in-complete. One potential approach is to generate the full-body motions from sparse inputs collected from limited sensors using a Neural Network (NN) model. In this paper, we propose a novel method based on a multi-layer perceptron (MLP) backbone that is enhanced with residual connections and a novel NN-component called Memory-Block. In particular, Memory-Block represents missing sensor data with trainable code-vectors, which are combined with the sparse signals from previous time instances to improve the temporal consistency. Furthermore, we formulate our solution as a multi-task learning problem, allowing our MLP-backbone to learn robust representations that boost accuracy. Our experiments show that our method outperforms state-of-the-art baselines by substantially reducing prediction errors. Moreover, it achieves 72 FPS on mobile HMDs that ultimately improves the accuracy-running time tradeoff.