Article
Geosciences, Multidisciplinary
Suihong Song, Tapan Mukerji, Jiagen Hou
Summary: Conditional facies modeling combines geological spatial patterns with different types of observed data to build earth models for predictions of subsurface resources. Researchers have improved GANs for conditional facies simulation by introducing an extra condition-based loss function and adjusting the architecture of the generator, resulting in high-quality facies models and stronger conditioning ability of the generators.
MATHEMATICAL GEOSCIENCES
(2021)
Article
Geosciences, Multidisciplinary
Fleford Redoloza, Liangping Li, Arden Davis
Summary: Groundwater-flow and contaminant-transport modeling can benefit from the use of machine learning techniques such as generative adversarial networks (GANs). This study focuses on a progressive growing GAN (PGGAN) to generate geologically realistic images of channel aquifers based on field observations. The conditioning behavior and its influence on the network architecture were measured using a novel metric called the conditioning ratio. The results revealed different conditioning behaviors based on the number of conditioning arrays injected into the generator.
HYDROGEOLOGY JOURNAL
(2023)
Article
Computer Science, Artificial Intelligence
Xiaoning Song, Yao Chen, Zhen-Hua Feng, Guosheng Hu, Dong-Jun Yu, Xiao-Jun Wu
Summary: The article presents a new Self-growing and Pruning Generative Adversarial Network (SP-GAN) for realistic image generation, which dynamically adjusts the network size and architecture, and utilizes an adaptive loss function. Experimental results show the method's advantages in stability and efficiency.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Renhao Sun, Yang Wang, Zhao Zhang, Richang Hong, Meng Wang
Summary: The study introduces a novel deep adversarial inconsistent cognitive sampling (DAICS) method for multiview clustering, which jointly learns a binary classifier and a deep consistent feature embedding network through an adversarial minimax game over difficulty labels of multiview consistent samples, enhancing clustering efficiency.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Keewon Shin, Jung Su Lee, Ji Young Lee, Hyunsu Lee, Jeongseok Kim, Jeong-Sik Byeon, Hwoon-Yong Jung, Do Hoon Kim, Namkug Kim
Summary: Generative adversarial networks (GAN) in medicine are valuable techniques for producing high-quality gastrointestinal images. This study used the progressive growing of GAN (PGGAN) to generate realistic images and investigated its limitations. The accuracy and sensitivity of endoscopists in distinguishing real and synthetic images were not significantly different. Real images with the anatomical landmark pylorus had higher detection sensitivity. However, GANs need improvement in representing rugal folds and mucous membrane texture.
JOURNAL OF DIGITAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Peng Zhou, Lingxi Xie, Bingbing Ni, Qi Tian
Summary: This article presents CIPS3D++, an upgraded version of style-based GANs aiming at high-robust, high-resolution, and high-efficiency 3D-aware image generation. It introduces CIPS-3D as the basic model with rotation-invariance and robustness and builds upon it with geometric regularization and upsampling operations to achieve high-resolution and high-quality image generation and editing.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Eduardo Perez, Sebastian Ventura
Summary: Early melanoma diagnosis is crucial for skin cancer treatment and can reduce mortality rates. Generative Adversarial Networks have been used to improve diagnostic capacity, but their application is challenging due to image variance, limited data, and model instability.
ARTIFICIAL INTELLIGENCE IN MEDICINE
(2023)
Article
Computer Science, Information Systems
Haixu Yang, Jihong Liu, Lvheng Zhang, Yan Li, Henggui Zhang
Summary: This study proposed a ProGAN-based ECG sample generation model, ProEGAN-MS, to address data imbalance issues, demonstrating higher fidelity and diversity of the generated data compared to other GAN-based ECG augmentation methods.
Article
Computer Science, Artificial Intelligence
Liangjun Chen, Zhengwang Wu, Dan Hu, Fan Wang, J. Keith Smith, Weili Lin, Li Wang, Dinggang Shen, Gang Li
Summary: Automatic correction of intensity nonuniformity is crucial in brain MR image analysis, especially for infant brain MR images. The proposed 3D adversarial bias correction network (ABCnet) is tailored for direct prediction of bias fields to handle regionally-heterogeneous intensity changes, showing superior performance in both accuracy and efficiency compared to existing methods.
MEDICAL IMAGE ANALYSIS
(2021)
Review
Remote Sensing
Shahab Jozdani, Dongmei Chen, Darren Pouliot, Brian Alan Johnson
Summary: Researchers conducted a comprehensive review and meta-analysis of the application of Generative Adversarial Networks (GANs) in the field of Remote Sensing (RS). They found that image classification, especially in urban mapping, has been the most popular application of GANs in RS. However, the potential of GANs for analyzing medium spatial-resolution multi-spectral images has not been fully explored. Furthermore, there are unresolved questions regarding the suitability of different GAN models for various applications and the extent to which GANs can replace real RS data.
INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION
(2022)
Article
Automation & Control Systems
Tong Zhang, Jingxiang Lian, Jingtao Wen, C. L. Philip Chen
Summary: Recent studies propose a novel top-down convolutional network for human pose estimation, which considers physical constraints and internal relationships of body parts. The network incorporates prior knowledge and uses adversarial learning to improve robustness in complex field conditions. Experimental results demonstrate that the proposed approach outperforms the original method and generates robust pose predictions on the MS COCO dataset.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Review
Computer Science, Theory & Methods
Eoin Brophy, Zhengwei Wang, Qi She, Tomas Ward
Summary: This article reviews the variants of generative adversarial networks (GANs) designed for time series related applications. It proposes a classification of discrete-variant GANs and continuous-variant GANs, showcasing the latest literature, architectures, results, and applications in this field. The article also covers evaluation metrics, privacy measures, and future directions for dealing with sensitive data.
ACM COMPUTING SURVEYS
(2023)
Article
Computer Science, Artificial Intelligence
Heechul Lim, Kang-Wook Chon, Min-Soo Kim
Summary: In supervised computer vision tasks, CNNs have proven to be superior to alternative methods. However, generating large-scale labeled datasets for training and validating these models is costly and requires specific expert knowledge. Active learning is a promising approach to generate labeled datasets with limited labeling budgets. This study addresses the issue of distractor points in active learning, proposing a method that effectively eliminates distractor points using a combination of existing active learning methods and a training strategy with GANs.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Interdisciplinary Applications
Francis Baek, Daeho Kim, Somin Park, Hyoungkwan Kim, SangHyun Lee
Summary: This study proposes a data augmentation method that combines generative adversarial networks with a target classifier to address the issue of data shortage in computer vision applications for construction. The results demonstrate that the proposed method significantly improves classification accuracy and enhances feature extraction for the target object.
JOURNAL OF COMPUTING IN CIVIL ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Jiamin Liang, Xin Yang, Yuhao Huang, Haoming Li, Shuangchi He, Xindi Hu, Zejian Chen, Wufeng Xue, Jun Cheng, Dong Ni
Summary: This paper proposes a generative adversarial network (GAN) based image synthesis framework for generating realistic and high-resolution B-mode ultrasound (US) images. The framework incorporates auxiliary sketch guidance and a progressive training strategy to enhance structural details and resolution of the generated images. The method is versatile and has been validated on multiple US image datasets of different anatomical structures.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Geosciences, Multidisciplinary
Alex Miltenberger, Tapan Mukerji, Jayaram Hariharan, Paola Passalacqua, Erik Nesvold
Summary: This study proposes a probabilistic framework based on Monte Carlo and metrics to test the ability of numerical delta models in capturing the link between surface dynamics and subsurface structures. The results show that certain delta surface features are informative of the spatial organization of sediment in the subsurface, while others are less informative. Key differences between experimental and numerical surface dynamics are likely due to limitations in numerical model resolution and assumptions in the model physics.
MATHEMATICAL GEOSCIENCES
(2022)
Article
Energy & Fuels
Sergei Petrov, Tapan Mukerji, Xin Zhang, Xinfei Yan
Summary: The task of seismic data interpretation is time-consuming and uncertain, but machine learning tools can provide a shortcut between raw seismic data and reservoir characteristics. Convolutional neural networks are efficient for seismic facies classification and interpretation. In this study, we experimented with three different convolutional architectures and compared their results and computational efficiency using synthetic and field datasets.
Article
Geochemistry & Geophysics
Michael Pimentel-Galvan, Kimberly V. Lau, Katharine Maher, Tapan Mukerji, Daniel J. Lehrmann, Demir Altiner, Jonathan L. Payne
Summary: This study investigates the uranium cycle across the Permian-Triassic boundary and the early Triassic using the Monte Carlo method and principal component analysis. The best-fitting models suggest a significant increase in seafloor anoxia during the Permian/Triassic transition, lasting from 20 kyr to 1.2 Myr. The extent and duration of anoxia show an inverse relationship, and there is no indication of complete re-oxygenation during the study interval.
GEOCHEMISTRY GEOPHYSICS GEOSYSTEMS
(2022)
Article
Geochemistry & Geophysics
Anshuman Pradhan, Tapan Mukerji
Summary: Deep learning applications in seismic reservoir characterization often require synthetic data generation. The article discusses a method for generating synthetic training data and highlights practical issues when training models on synthetic seismic data, using a real case study as an example.
Article
Geochemistry & Geophysics
Suihong Song, Tapan Mukerji, Jiagen Hou
Summary: This study improves the simulation method based on generative adversarial networks (GANs) to bridge the gap between remotely sensed geophysical information and geology. The generated geological facies models are realistic, diversified, and consistent with all input conditions.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Tongjun Chen, Zhijun Lin, Zuiliang Liu, Tapan Mukerji
Summary: This study conducted a comparative experiment to discuss the detectability and influencing factors of heterogeneous stress field distributions in underground coal mines in North China. The results showed that curvature variations and mined-out voids are the dominant factors affecting the uneven distribution of stress fields in coal mine panels.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geosciences, Multidisciplinary
Runhai Feng, Dario Grana, Tapan Mukerji, Klaus Mosegaard
Summary: Geological facies modeling is a key component in exploring and characterizing subsurface reservoirs. This work presents a deep learning approach based on generative adversarial networks for geological facies modeling. It introduces a Bayesian GANs approach to create facies models and analyze the model uncertainty. The proposed method is applied to different geological scenarios and successfully captures the variability of the data.
MATHEMATICAL GEOSCIENCES
(2022)
Article
Geosciences, Multidisciplinary
Mingliang Liu, Tapan Mukerji
Summary: This paper introduces a method based on deep generative adversarial networks to generate high-resolution digital rock images and recover fine details. Experimental results demonstrate the effectiveness of this method and its potential to better characterize heterogeneous porous media and predict pore-scale flow and petrophysical properties.
GEOPHYSICAL RESEARCH LETTERS
(2022)
Article
Geochemistry & Geophysics
Liming Zhao, Tongjun Chen, Tapan Mukerji, Mingjin Zhang, Tao Xing
Summary: This article proposes a model based on Berryman and Milton's generalized Gassmann's equations to calculate the high-frequency saturated bulk modulus at different soft-pore fractions or crack densities in Mavko and Jizba's model. Experimental data validate the effectiveness of the proposed model.
Article
Geosciences, Multidisciplinary
Mingliang Liu, Dario Grana, Tapan Mukerji
Summary: A computationally efficient method using randomized tensor decomposition is developed in this study to reduce model parameters and observations for efficient data assimilation in low-dimensional spaces.
MATHEMATICAL GEOSCIENCES
(2022)
Article
Computer Science, Interdisciplinary Applications
Ali Kashefi, Tapan Mukerji
Summary: We propose a novel physics-informed deep learning framework for solving steady-state incompressible flow on multiple sets of irregular geometries. This framework incorporates a point-cloud based neural network to capture the geometric features of computational domains and utilizes the mean squared residuals of the governing partial differential equations as the loss function to capture the physics. It allows for solving equations on a set of computational domains with irregular geometries and can predict solutions on domains with unseen geometries, resulting in cost savings.
JOURNAL OF COMPUTATIONAL PHYSICS
(2022)
Article
Engineering, Civil
Suihong Song, Dongxiao Zhang, Tapan Mukerji, Nanzhe Wang
Summary: To address the challenging task of stochastic conditional geomodelling, we propose a deep-learning framework called GANSim-surrogate, which effectively integrates geological patterns and various types of data. The framework consists of a CNN generator, a CNN-based surrogate, and options for searching appropriate input latent vectors. Through validation on channelized reservoirs, the framework is proven to generate realistic and consistent models with all conditioning data, while also being computationally efficient.
JOURNAL OF HYDROLOGY
(2023)
Article
Materials Science, Multidisciplinary
Rasool Ahmad, Mingliang Liu, Michael Ortiz, Tapan Mukerji, Wei Cai
Summary: This study focuses on calculating the homogenized elastic properties of rocks using 3D micro-CT scanned images. To solve the problem of large micro-CT images, a hierarchical homogenization method is proposed, where the image is divided into smaller subimages. The subimages are individually homogenized and then assembled to find the final homogenized elastic constant. The error in the homogenized constant follows a power law scaling with respect to the subimage size, and this scaling is used for better approximation of large heterogeneous microstructures.
JOURNAL OF THE MECHANICS AND PHYSICS OF SOLIDS
(2023)
Article
Geochemistry & Geophysics
Mingliang Liu, Divakar Vashisth, Dario Grana, Tapan Mukerji
Summary: A differentiable physics model integrated with neural networks is developed for high-resolution reservoir monitoring. The proposed method effectively estimates reservoir properties and accurately quantifies uncertain parameters for CO2 storage monitoring.
JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH
(2023)
Article
Geochemistry & Geophysics
Mingliang Liu, Rasool Ahmad, Wei Cai, Tapan Mukerji
Summary: Digital rock physics combines tomographic imaging techniques with numerical simulations to estimate effective rock properties. To address the computational challenge of large sample sizes, a hierarchical homogenization method with a data-driven surrogate model based on convolutional neural networks is proposed. The method reduces computational time and memory demand compared to conventional algorithms.
JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH
(2023)