MME: Mixture of Mesh Experts with Random Walk Transformer Gating

2026-02-28

AI Summary

This work proposes a novel Mixture-of-Experts (MoE)–based fusion framework to address the limited generalization of existing mesh analysis methods across diverse object categories. The approach introduces an innovative Transformer gating mechanism driven by random walks on mesh surfaces, enabling each expert to specialize in its most competent category. Coupled with an attention mechanism, the model dynamically selects the most discriminative regions for decision-making. During training, a dynamic loss balancing strategy harmonizes diversity and similarity objectives, significantly enhancing the model’s representational capacity. Extensive experiments demonstrate that the proposed method achieves state-of-the-art performance across three fundamental tasks—mesh classification, retrieval, and semantic segmentation—consistently outperforming current approaches by a notable margin.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Multi-modal Vision
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

In recent years, various methods have been proposed for mesh analysis, each offering distinct advantages and often excelling on different object classes. We present a novel Mixture of Experts (MoE) framework designed to harness the complementary strengths of these diverse approaches. We propose a new gate architecture that encourages each expert to specialise in the classes it excels in. Our design is guided by two key ideas: (1) random walks over the mesh surface effectively capture the regions that individual experts attend to, and (2) an attention mechanism that enables the gate to focus on the areas most informative for each expert's decision-making. To further enhance performance, we introduce a dynamic loss balancing scheme that adjusts a trade-off between diversity and similarity losses throughout the training, where diversity prompts expert specialization, and similarity enables knowledge sharing among the experts. Our framework achieves state-of-the-art results in mesh classification, retrieval, and semantic segmentation tasks. Our code is available at: https://github.com/amirbelder/MME-Mixture-of-Mesh-Experts.

Problem

• mesh analysis
• Mixture of Experts
• expert specialization
• 3D mesh processing
• multi-expert integration

Innovation

• Mixture of Experts
• Random Walk Transformer
• Mesh Analysis
• Dynamic Loss Balancing
• Attention Mechanism