Neural Clustering for Prefractured Mesh Generation in Real-time Object Destruction

2024-12-02

SIGGRAPH Asia Posters

AI Summary

Pre-fracture mesh generation for real-time object destruction simulation traditionally relies on heuristic clustering methods, leading to physically implausible deformations and structurally inconsistent fragmentation. Method: This paper reformulates pre-fracture clustering as an unordered point cloud segmentation task and proposes the first physics-simulation-data-driven, end-to-end neural clustering paradigm. We construct a large-scale point cloud dataset annotated with structural fragility labels and train a deep neural network to directly predict internal weak regions of objects, enabling high-fidelity pre-fracture mesh generation. Contribution/Results: Our approach overcomes the representational limitations of conventional heuristic clustering. While preserving real-time performance, it significantly improves both visual authenticity and physical plausibility of destruction sequences. The framework establishes a learnable, generalizable structural analysis paradigm for interactive destruction simulation.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Clustering
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Prefracture method is a practical implementation for real-time object destruction that is hardly achievable within performance constraints, but can produce unrealistic results due to its heuristic nature. To mitigate it, we approach the clustering of prefractured mesh generation as an unordered segmentation on point cloud data, and propose leveraging the deep neural network trained on a physics-based dataset. Our novel paradigm successfully predicts the structural weakness of object that have been limited, exhibiting ready-to-use results with remarkable quality.

Problem

• Neural clustering for prefracture mesh generation
• Real-time object destruction with structural weakness prediction
• Physics-based dataset training for realistic results

Innovation

• Neural clustering for mesh generation
• Physics-based deep neural network
• Real-time object destruction simulation