Vision Mamba for Permeability Prediction of Porous Media

2025-10-16

AI Summary

This study addresses the high computational cost, large parameter count, and poor scalability of existing models for predicting permeability in 3D porous media. We propose the first end-to-end regression framework leveraging Vision Mamba—a state-of-the-art linear-complexity vision architecture—integrated with 3D voxel encoding to efficiently represent pore-structure data. Experiments demonstrate that our model outperforms ViT and CNN baselines in MAE and RMSE, achieves a 2.1× inference speedup, and reduces GPU memory consumption by 37%. Our key contributions are: (1) the pioneering adaptation of Vision Mamba to physical property prediction in porous media; (2) empirical validation of its effectiveness and efficiency in low-data-regime, high-resolution 3D structural modeling; and (3) open-sourcing of training code and standardized data preprocessing pipelines to ensure reproducibility.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Vision Mamba has recently received attention as an alternative to Vision Transformers (ViTs) for image classification. The network size of Vision Mamba scales linearly with input image resolution, whereas ViTs scale quadratically, a feature that improves computational and memory efficiency. Moreover, Vision Mamba requires a significantly smaller number of trainable parameters than traditional convolutional neural networks (CNNs), and thus, they can be more memory efficient. Because of these features, we introduce, for the first time, a neural network that uses Vision Mamba as its backbone for predicting the permeability of three-dimensional porous media. We compare the performance of Vision Mamba with ViT and CNN models across multiple aspects of permeability prediction and perform an ablation study to assess the effects of its components on accuracy. We demonstrate in practice the aforementioned advantages of Vision Mamba over ViTs and CNNs in the permeability prediction of three-dimensional porous media. We make the source code publicly available to facilitate reproducibility and to enable other researchers to build on and extend this work. We believe the proposed framework has the potential to be integrated into large vision models in which Vision Mamba is used instead of ViTs.

Problem

• Predicting permeability of 3D porous media using Vision Mamba
• Comparing Vision Mamba with ViT and CNN models for permeability prediction
• Demonstrating computational efficiency advantages over traditional vision models

Innovation

• Vision Mamba backbone for permeability prediction
• Linear scaling network size with image resolution
• Fewer trainable parameters than CNNs and ViTs