A Locality-based Neural Solver for Optical Motion Capture

TL;DR

This paper introduces a locality-based graph neural network approach for cleaning and solving noisy optical motion capture data, effectively handling occlusions and tracking errors to improve motion reconstruction accuracy.

Contribution

The paper presents a novel heterogeneous graph neural network that leverages local marker correlations and a masking-based training regime for enhanced motion capture data cleaning.

Findings

Outperforms state-of-the-art in marker position error by ~20%.

Reduces joint reconstruction errors by 30%.

Effectively handles occlusions and tracking errors.

Abstract

We present a novel locality-based learning method for cleaning and solving optical motion capture data. Given noisy marker data, we propose a new heterogeneous graph neural network which treats markers and joints as different types of nodes, and uses graph convolution operations to extract the local features of markers and joints and transform them to clean motions. To deal with anomaly markers (e.g. occluded or with big tracking errors), the key insight is that a marker's motion shows strong correlations with the motions of its immediate neighboring markers but less so with other markers, a.k.a. locality, which enables us to efficiently fill missing markers (e.g. due to occlusion). Additionally, we also identify marker outliers due to tracking errors by investigating their acceleration profiles. Finally, we propose a training regime based on representation learning and data augmentation, by training the model on data with masking. The masking schemes aim to mimic the occluded and noisy markers often observed in the real data. Finally, we show that our method achieves high accuracy on multiple metrics across various datasets. Extensive comparison shows our method outperforms state-of-the-art methods in terms of prediction accuracy of occluded marker position error by approximately 20%, which leads to a further error reduction on the reconstructed joint rotations and positions by 30%. The code and data for this paper are available at https://github.com/non-void/LocalMoCap.