NeuralPVS: Learned Estimation of Potentially Visible Sets

2025-09-29

AI Summary

Real-time visibility computation has long suffered from high computational overhead and limitations of static precomputation in large-scale or dynamic scenes. To address this, we propose the first deep learning–based approach for visible set prediction, introducing an end-to-end learning framework grounded in region-wise visibility. Our method employs a 3D volume-preserving interleaved compression architecture to efficiently encode spatial relationships, and introduces a repulsive visibility loss function to enhance convergence, generalization, and robustness. We represent scenes via voxelization and extract geometric features using sparse convolutional networks. The system achieves real-time inference at 100 Hz with a geometric omission rate below 1%, significantly outperforming state-of-the-art methods in both accuracy and efficiency. This work establishes a new paradigm for high-fidelity, real-time visibility prediction.

Technology Category

• Computer Vision: 3D Computer Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Learning with Manifolds

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Security and Privacy: Data transparency and provenance

Abstract

Real-time visibility determination in expansive or dynamically changing environments has long posed a significant challenge in computer graphics. Existing techniques are computationally expensive and often applied as a precomputation step on a static scene. We present NeuralPVS, the first deep-learning approach for visibility computation that efficiently determines from-region visibility in a large scene, running at approximately 100 Hz processing with less than $1%$ missing geometry. This approach is possible by using a neural network operating on a voxelized representation of the scene. The network's performance is achieved by combining sparse convolution with a 3D volume-preserving interleaving for data compression. Moreover, we introduce a novel repulsive visibility loss that can effectively guide the network to converge to the correct data distribution. This loss provides enhanced robustness and generalization to unseen scenes. Our results demonstrate that NeuralPVS outperforms existing methods in terms of both accuracy and efficiency, making it a promising solution for real-time visibility computation.

Problem

• Real-time visibility computation in dynamic large-scale environments
• Overcoming computational expense of existing precomputation-based methods
• Achieving high accuracy with neural network-based visibility estimation

Innovation

• Neural network operates on voxelized scene representation
• Combines sparse convolution with 3D volume-preserving interleaving
• Introduces novel repulsive visibility loss for convergence