Progress and Challenges in Deep Learning Analysis of Geoscience Images

Lima, Rafael Pires de; Marfurt, Kurt J.; Duarte, D.; Bonar, Alicia

doi:10.3997/2214-4609.201901607

Cited by 11 publications

(4 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Scene classification is a fundamental remote-sensing task and important for many practical remote-sensing applications, such as urban planning [4], land management [5], and to characterize wild fires [6,7], among other applications. Such ample use of remote-sensing image of porosity in thin section images, and Pires de Lima et al [31,32] for the classification of a variety of geoscience images. Minaee et al [33] stated that many of the deep neural network models for biometric recognition are based on transfer learning.…”

Section: Introductionmentioning

confidence: 99%

Convolutional Neural Network for Remote-Sensing Scene Classification: Transfer Learning Analysis

2019

View full text Add to dashboard Cite

Remote-sensing image scene classification can provide significant value, ranging from forest fire monitoring to land-use and land-cover classification. Beginning with the first aerial photographs of the early 20th century to the satellite imagery of today, the amount of remote-sensing data has increased geometrically with a higher resolution. The need to analyze these modern digital data motivated research to accelerate remote-sensing image classification. Fortunately, great advances have been made by the computer vision community to classify natural images or photographs taken with an ordinary camera. Natural image datasets can range up to millions of samples and are, therefore, amenable to deep-learning techniques. Many fields of science, remote sensing included, were able to exploit the success of natural image classification by convolutional neural network models using a technique commonly called transfer learning. We provide a systematic review of transfer learning application for scene classification using different datasets and different deep-learning models. We evaluate how the specialization of convolutional neural network models affects the transfer learning process by splitting original models in different points. As expected, we find the choice of hyperparameters used to train the model has a significant influence on the final performance of the models. Curiously, we find transfer learning from models trained on larger, more generic natural images datasets outperformed transfer learning from models trained directly on smaller remotely sensed datasets. Nonetheless, results show that transfer learning provides a powerful tool for remote-sensing scene classification.

show abstract

Section: Introductionmentioning

confidence: 99%

Convolutional Neural Network for Remote-Sensing Scene Classification: Transfer Learning Analysis

2019

View full text Add to dashboard Cite

show abstract

“…ML techniques have proven a successful range of applications in various domains of subsurface engineering, including but not limited to petrophysics and well logging, production, well testing, exploration, rock characterization, and digital rock imaging. − Other ML techniques, such as GBR, support vector regression, and tree-based approaches, have also been successfully applied in various domains of subsurface engineering. ,− …”

Section: Predictive Modeling Designmentioning

confidence: 99%

An Experimental Study and Machine Learning Modeling of Shale Swelling in Extended Reach Wells When Exposed to Diverse Water-Based Drilling Fluids

Tariq,

Murtaza,

Alrasheed

et al. 2024

Energy Fuels

View full text Add to dashboard Cite

Shale swelling poses considerable challenges for companies involved in extended-reach well drilling, particularly when it comes to maintaining wellbore stability. Despite the incorporation of swelling inhibitors in water-based drilling fluids (WBDFs), shale swelling can persist as an issue. Currently, the assessment of WBDFs' ability to prevent swelling usually involves costly, timeconsuming, and labor-intensive laboratory experiments. Therefore, this study leverages machine learning techniques (ML) to forecast the dynamic linear swelling behavior of sodium bentonite-based shale wafers. These shale wafers were subjected to different WBDFs containing diverse inorganic salts such as sodium chloride (NaCl), potassium chloride (KCl), magnesium chloride (MgCl 2 ), and calcium chloride (CaCl 2 ). To gather sufficient data to train the ML models, an extensive experimental study was conducted using various WBDF formulations. The experiments were conducted on wafers, wherein each wafer treated with an inorganic salt underwent a linear swell test for 120 h until the swelling reached a plateau. Moreover, the conductivities and zeta potential of WBDFs, which were prepared using varying concentrations of salts, were recorded. Several ML techniques, namely gradient boosting, decision trees, adaptive gradient boosting (AdaBoost), K-nearest neighbors, random forest, extreme gradient boosting, and stacked generalized regression (SGR) were used to forecast shale swelling. The ML models were trained using input features such as salt types, salt concentrations, salt conductivities, salt zeta potential, and elapsed time. The results revealed that the SGR model outperformed other techniques by effectively predicting linear swelling in terms of coefficient of determination (R 2 ) above 0.95. The developed ML model offers an efficient approach to assess the maximum swelling potential of various WBDFs that can help in effectively mitigating the wellbore instability concerns by continuously evaluating the interaction between shale and the drilling fluid for drilling for extended reach or maximum contact wells.

show abstract

“…However, their analysis primarily caters to the geoscience community, lacking a detailed exploration of DLbased techniques. Similarly, a survey by De et al [51] provides insightful analysis of fossil classification via convolutional neural networks (CNNs). This survey is limited because it does not cover a broader spectrum of DL architectures, particularly recent advancements such as transformers and generative models.…”

Section: Introductionmentioning

confidence: 99%

Advancing Paleontology: A Survey on Deep Learning Methodologies in Fossil Image Analysis

Ansari,

Ishaq,

Yusuf

et al. 2024

Preprint

View full text Add to dashboard Cite

Understanding ancient organisms and their paleo-environments through the study of body fossils represents a central tenet of paleontology. Advances in digital image capture over the past several decades now allow for efficient and accurate documentation, curation and interrogation of fossil anatomy over disparate length scales. Despite these developments, key body fossil image processing and analysis tasks, such as segmentation and classification still require significant user intervention, which can be labor-intensive and subject to human bias. Recent advancements in deep learning offer the potential to automate fossil image analysis while improving throughput and limiting operator bias. Despite the recent emergence of deep learning within paleontology, challenges such as the scarcity of diverse, high quality image datasets and the complexity of fossil morphology necessitate further advancements and the adoption of concepts from other scientific domains. Here, we comprehensively review state-of-the-art deep learning-based methodologies applied towards body fossil analysis while grouping the studies based on the fossil type and nature of the task. Furthermore, we analyze existing literature to tabulate dataset information, neural network architecture type, key results, and comprehensive textual summaries. Finally, based on the collective limitations of the existing studies, we discuss novel techniques for fossil data augmentation and fossil image enhancements, which can be combined with advanced neural network architectures, such as diffusion models, generative hybrid networks, transformers, and graph neural networks, to improve body fossil image analysis.

show abstract

Progress and Challenges in Deep Learning Analysis of Geoscience Images

Cited by 11 publications

References 10 publications

Convolutional Neural Network for Remote-Sensing Scene Classification: Transfer Learning Analysis

Convolutional Neural Network for Remote-Sensing Scene Classification: Transfer Learning Analysis

An Experimental Study and Machine Learning Modeling of Shale Swelling in Extended Reach Wells When Exposed to Diverse Water-Based Drilling Fluids

Advancing Paleontology: A Survey on Deep Learning Methodologies in Fossil Image Analysis

Contact Info

Product

Resources

About