E2E-GNet: An End-to-End Skeleton-based Geometric Deep Neural Network for Human Motion Recognition

2026-03-02

AI Summary

This work proposes E2E-GNet, an end-to-end geometric deep neural network designed to address the challenges of modeling skeleton sequences in non-Euclidean spaces. By integrating geometric transformation layers with differentiable logarithmic map activations, the method projects skeleton motion sequences from non-Euclidean manifolds into a linear space. It further introduces, for the first time, a distortion-aware optimization mechanism that effectively preserves essential geometric structures during joint optimization. Combining geometric deep learning, differentiable logarithmic mapping, and distortion-aware optimization, E2E-GNet enables efficient end-to-end modeling of skeleton sequences. Extensive experiments on five cross-domain datasets demonstrate that the proposed approach significantly outperforms existing methods, achieving higher action recognition accuracy while reducing computational overhead.

Technology Category

• Machine Learning: Learning with Manifolds
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Computer Vision: Motion & Tracking

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

Geometric deep learning has recently gained significant attention in the computer vision community for its ability to capture meaningful representations of data lying in a non-Euclidean space. To this end, we propose E2E-GNet, an end-to-end geometric deep neural network for skeleton-based human motion recognition. To enhance the discriminative power between different motions in the non-Euclidean space, E2E-GNet introduces a geometric transformation layer that jointly optimizes skeleton motion sequences on this space and applies a differentiable logarithm map activation to project them onto a linear space. Building on this, we further design a distortion-aware optimization layer that limits skeleton shape distortions caused by this projection, enabling the network to retain discriminative geometric cues and achieve a higher motion recognition rate. We demonstrate the impact of each layer through ablation studies and extensive experiments across five datasets spanning three domains show that E2E-GNet outperforms other methods with lower cost.

Problem

• human motion recognition
• skeleton-based
• geometric deep learning
• non-Euclidean space
• end-to-end

Innovation

• geometric deep learning
• skeleton-based motion recognition
• logarithm map activation
• distortion-aware optimization
• non-Euclidean space