Application of Graph Convolutions in a Lightweight Model for Skeletal Human Motion Forecasting
This work addresses motion prediction for human-robot interaction, but appears incremental as it applies existing graph convolutions to a known problem.
The paper tackles human motion forecasting by exploiting skeletal structure through graph convolutions, achieving competitive predictions with a lightweight model requiring fewer parameters.
Prediction of movements is essential for successful cooperation with intelligent systems. We propose a model that integrates organized spatial information as given through the moving body's skeletal structure. This inherent structure is exploited in our model through application of Graph Convolutions and we demonstrate how this allows leveraging the structured spatial information into competitive predictions that are based on a lightweight model that requires a comparatively small number of parameters.