Image Augmentation Techniques Research Articles

LCAMix: Local-and-contour aware grid mixing based data augmentation for medical image segmentation

Medical image segmentation often faces challenges related to overfitting, primarily due to the limited and complex training samples. This challenge often prompts the use of self-supervised learning and data augmentation. However, self-supervised learning requires well-defined hand-crafted tasks and multiple training stages. On the other hand, basic image augmentation techniques like cropping, rotation, and flipping, effective for natural scene images, have limited efficacy for medical images due to their isotropic nature.While regional dropout regularization data augmentation methods have proven effective in image recognition tasks, their application in image segmentation is not as extensively studied. Additionally, existing augmentation methods often operate on square regions, leading to the loss of crucial contour information. This is particularly problematic for medical image segmentation tasks dealing with regions of interest characterized by intricate shapes. In this work, we introduce LCAMix, a novel data augmentation approach designed for medical image segmentation. LCAMix operates by blending two images and their segmentation masks based on their superpixels, incorporating a local-and-contour-aware strategy. The training process on augmented images adopts two auxiliary pretext tasks: firstly, classifying local superpixels in augmented images using an adaptive focal margin, leveraging segmentation ground truth masks as prior knowledge; secondly, reconstructing the two source images using mixed superpixels as mutual masks, emphasizing spatial sensitivity. Our method stands out as a simple, one-stage, model-agnostic, and plug-and-play data augmentation solution applicable to various segmentation tasks. Notably, it requires no external data or additional models. Extensive experiments validate its superior performance across diverse medical segmentation datasets and tasks. The source codes are available at https://github.com/DanielaPlusPlus/DataAug4Medical.

A Hybrid Image Augmentation Technique for User- and Environment-Independent Hand Gesture Recognition Based on Deep Learning

This research stems from the increasing use of hand gestures in various applications, such as sign language recognition to electronic device control. The focus is the importance of accuracy and robustness in recognizing hand gestures to avoid misinterpretation and instruction errors. However, many experiments on hand gesture recognition are conducted in limited laboratory environments, which do not fully reflect the everyday use of hand gestures. Therefore, the importance of an ideal background in hand gesture recognition, involving only the signer without any distracting background, is highlighted. In the real world, the use of hand gestures involves various unique environmental conditions, including differences in background colors, varying lighting conditions, and different hand gesture positions. However, the datasets available to train hand gesture recognition models often lack sufficient variability, thereby hindering the development of accurate and adaptable systems. This research aims to develop a robust hand gesture recognition model capable of operating effectively in diverse real-world environments. By leveraging deep learning-based image augmentation techniques, the study seeks to enhance the accuracy of hand gesture recognition by simulating various environmental conditions. Through data duplication and augmentation methods, including background, geometric, and lighting adjustments, the diversity of the primary dataset is expanded to improve the effectiveness of model training. It is important to note that the utilization of the green screen technique, combined with geometric and lighting augmentation, significantly contributes to the model’s ability to recognize hand gestures accurately. The research results show a significant improvement in accuracy, especially with implementing the proposed green screen technique, underscoring its effectiveness in adapting to various environmental contexts. Additionally, the study emphasizes the importance of adjusting augmentation techniques to the dataset’s characteristics for optimal performance. These findings provide valuable insights into the practical application of hand gesture recognition technology and pave the way for further research in tailoring techniques to datasets with varying complexities and environmental variations.

A Comparison of CNN-based Image Feature Extractors for Weld Defects Classification

<span lang="EN-US">Classification of the types of weld defects is one of the stages of evaluating radiographic images, which is an essential step in controlling the quality of welded joints in materials. By automating the weld defects classification based on deep learning and the CNN architecture, it is possible to overcome the limitations of visually or manually evaluating radiographic images. Good accuracy in classification models for weld defects requires the availability of sufficient datasets. In reality, however, the radiographic image dataset accessible to the public is limited and imbalanced between classes. Consequently, simple image cropping and augmentation techniques are implemented during the data preparation stage. To construct a weld defect classification model, we proposed to utilize the transfer learning method by employing a pre-trained CNN architecture as a feature extractor, including DenseNet201, InceptionV3, MobileNetV2, NASNetMobile, ResNet50V2, VGG16, VGG19, and Xception, which are linked to a simple classification model based on multilayer perceptron. The test results indicate that the three best classification models were obtained by using the DenseNet201 feature extractor with a test accuracy value of 100%, followed by ResNet50V2 and InceptionV3 with an accuracy of 99.17%. These outcomes are better compared to state-of-the-art classification models with a maximum of six classes of defects. The research findings may assist radiography experts in evaluating radiographic images more accurately and efficiently.</span>

Automated Reservoir Characterization of Carbonate Rocks using Deep Learning Image Segmentation Approach

Summary The objective of this study is to develop a systematic and novel workflow for the automated and objective characterization of carbonate reservoirs with the help of deep learning architectures. An image database of more than 6,000 carbonate thin-section images was generated using the optical microscope and image augmentation techniques. Five features, namely clay/silt/mineral, calcite, pores, fossils, and opaque minerals, were identified with the help of manual petrography of the thin sections under the microscope. A total of four deep learning models were developed, which included U-Net, U-Net with ResNet34 backbone, U-Net with Mobilenetv2 backbone, and LinkNet with ResNet34 backbone. The Ensemble model of U-Net + ResNet34 and U-Net + MobileNetv2 yielded the highest intersection over union (IoU) score of 75%, followed by the U-Net + ResNet34 model with an IoU score of 61%. The models struggled with class imbalance, which was very prominent in the image database, with classes such as fossils and opaques considered to be rare. The statistical analysis of the relative errors revealed that the major classes play a more important role in increasing the final IoU score as opposed to the common understanding that the rare classes affect the model performance. The novel workflow developed in this paper can be extended to real carbonate reservoirs for time efficient, objective, and accurate characterization.

Experiments on image data augmentation techniques for geological rock type classification with convolutional neural networks

The integration of image analysis through deep learning (DL) into rock classification represents a significant leap forward in geological research. While traditional methods remain invaluable for their expertise and historical context, DL offers a powerful complement by enhancing the speed, objectivity, and precision of the classification process. This research explores the significance of image data augmentation techniques in optimizing the performance of convolutional neural networks (CNNs) for geological image analysis, particularly in the classification of igneous, metamorphic, and sedimentary rock types from rock thin section (RTS) images. This study primarily focuses on classic image augmentation techniques and evaluates their impact on model accuracy and precision. Results demonstrate that augmentation techniques like Equalize significantly enhance the model's classification capabilities, achieving an F1-Score of 0.9869 for igneous rocks, 0.9884 for metamorphic rocks, and 0.9929 for sedimentary rocks, representing improvements compared to the baseline original results. Moreover, the weighted average F1-Score across all classes and techniques is 0.9886, indicating an enhancement. Conversely, methods like Distort lead to decreased accuracy and F1-Score, with an F1-Score of 0.949 for igneous rocks, 0.954 for metamorphic rocks, and 0.9416 for sedimentary rocks, exacerbating the performance compared to the baseline. The study underscores the practicality of image data augmentation in geological image classification and advocates for the adoption of DL methods in this domain for automation and improved results. The findings of this study can benefit various fields, including remote sensing, mineral exploration, and environmental monitoring, by enhancing the accuracy of geological image analysis both for scientific research and industrial applications.

Investigating Training Datasets of Real and Synthetic Images for Outdoor Swimmer Localisation with YOLO

In this study, we developed and explored a methodical image augmentation technique for swimmer localisation in northern German outdoor lake environments. When it comes to enhancing swimmer safety, a main issue we have to deal with is the lack of real-world training data of such outdoor environments. Natural lighting changes, dynamic water textures, and barely visible swimming persons are key issues to address. We account for these difficulties by adopting an effective background removal technique with available training data. This allows us to edit swimmers into natural environment backgrounds for use in subsequent image augmentation. We created 17 training datasets with real images, synthetic images, and a mixture of both to investigate different aspects and characteristics of the proposed approach. The datasets were used to train YOLO architectures for possible future applications in real-time detection. The trained frameworks were then tested and evaluated on outdoor environment imagery acquired using a safety drone to investigate and confirm their usefulness for outdoor swimmer localisation.

A novel hybrid deep CNN model for breast cancer classification using Lipschitz-based image augmentation and recursive feature elimination

Breast cancer is the leading cause of death for women worldwide. Despite its difficulties, the histopathological diagnosis must be used because modern imaging techniques cannot differentiate between benign and malignant tumors. This article introduces a hybrid deep learning method with recursive feature elimination for categorizing histopathology images of breast lesions as benign or malignant, using standard datasets, to increase diagnostic accuracy. The proposed four-stage model includes several novel techniques. First, a modified Lipschitz-based image augmentation technique is utilized to ensure a proportionate representation of various sample classes. K-fold cross-validation is employed to mitigate the bias that emerges from the selection of training and testing datasets. Then, for efficient feature extraction, a hybrid convolutional neural network (CNN) approach is used by integrating three CNNs (PResNet-34, FE-VGG-16, and M-AlexNet). With the proposed universum twin support vector machine-based recursive feature elimination technique, the initial 1,000 features extracted by each deep model are further reduced to 100. Finally, the deep features are subsequently combined into a set of 300 features, which are then fed into machine learning classifiers such as decision tree, k-nearest neighborhood, linear discriminant analysis, linear regression, and support vector machine. The results demonstrate that the proposed hybrid CNN model outperforms state-of-the-art techniques in accurately classifying breast lesions as benign or malignant, with an accuracy of 99.99 %, sensitivity of 99.56 %, specificity of 99.53 %, precision of 99.55 %, and F1 score of 99.77 %.

Improving efficiency and accuracy of levee hazard detection with deep learning

This paper presents a model framework based on a deep learning algorithm using Faster R-CNN to improve the accuracy and efficiency of detecting long-distance levee hazards. The proposed method aims to assist technicians in identifying hazards in resistivity inversion profiles, thereby reducing the manual workload and extracting valuable information from a large volume of data using deep learning techniques. To this end, we constructed a dataset of levee hazards using inversion images of electrical resistivity imaging data obtained from numerical simulations, indoor experiments, and actual levee hazards. Subsequently, we utilized image augmentation techniques to expand the dataset and enhance the model's performance, resulting in a total of 2000 labeled hazard images. The network was then trained from scratch by setting the hyperparameters. the trained weight parameters and network architecture were saved for future use. The model's performance was evaluated using standard evaluation metrics, including precision, recall, F1 score, and average precision (AP). The proposed model demonstrated robustness and adaptability, achieving high precision and accuracy with an AP value of up to 100%. Furthermore, we validated the method using three cases, and the results indicated its superiority over traditional manual identification-based or outlier analysis detection of levee hazards in practical scenarios. This study emphasizes the significant advantages of deep learning techniques over the traditional methods, including improved accuracy, automation, speed, and scalability. It suggests that intelligent detection based on deep learning can accurately identify and locate hazards in levees.

Optimizing Facial Expression Recognition with Image Augmentation Techniques: VGG19 Approach on FERC Dataset

In the field of facial expression recognition (FER), the availability of balanced and representative datasets is key to success in training accurate models. However, Facial Expression Recognition Challenge (FERC) datasets often face the challenge of class imbalance, where some facial expressions have a much smaller number of samples compared to others. This issue can result in biased and unsatisfactory model performance, especially in recognizing less common facial expressions. Data augmentation techniques are becoming an important strategy as they can expand the dataset by creating new variations of existing samples, thus increasing the variety and diversity of the data. Data augmentation can be used to increase the number of samples for less common facial expression classes, thus improving the model's ability to recognize and understand diverse facial expressions. The augmentation results are then combined with balancing techniques such as SMOTE coupled with undersampling to improve model performance. In this study, VGG19 is used to support better model performance. This will provide valuable guidelines for optimizing more advanced CNN models in the future and may encourage further research in creating more innovative augmentation techniques.

Boosting Vehicle Classification with Augmentation Techniques across Multiple YOLO Versions

In recent years, computer vision has experienced a surge in applications across various domains, including product and quality inspection, automatic surveillance, and robotics. This study proposes techniques to enhance vehicle object detection and classification using augmentation methods based on the YOLO (You Only Look Once) network. The primary objective of the trained model is to generate a local vehicle detection system for Malaysia which have the capacity to detect vehicles manufactured in Malaysia, adapt to the specific environmental factors in Malaysia, and accommodate varying lighting conditions prevalent in Malaysia. The dataset used for this paper to develop and evaluate the proposed system was provided by a highway company, which captured a comprehensive top-down view of the highway using a surveillance camera. Rigorous manual annotation was employed to ensure accurate annotations within the dataset. Various image augmentation techniques were also applied to enhance the dataset's diversity and improve the system's robustness. Experiments were conducted using different versions of the YOLO network, such as YOLOv5, YOLOv6, YOLOv7, and YOLOv8, each with varying hyperparameter settings. These experiments aimed to identify the optimal configuration for the given dataset. The experimental results demonstrated the superiority of YOLOv8 over other YOLO versions, achieving an impressive mean average precision of 97.9% for vehicle detection. Moreover, data augmentation effectively solves the issues of overfitting and data imbalance while providing diverse perspectives in the dataset. Future research can focus on optimizing computational efficiency for real-time applications and large-scale deployments.

Advancing Cough Classification: Swin Transformer vs. 2D CNN with STFT and Augmentation Techniques

Coughing, a common symptom associated with various respiratory problems, is a crucial indicator for diagnosing and tracking respiratory diseases. Accurate identification and categorization of cough sounds, specially distinguishing between wet and dry coughs, are essential for understanding underlying health conditions. This research focuses on applying the Swin Transformer for classifying wet and dry coughs using short-time Fourier transform (STFT) representations. We conduct a comprehensive evaluation, including a performance comparison with a 2D convolutional neural network (2D CNN) model, and exploration of two distinct image augmentation methods: time mask augmentation and classical image augmentation techniques. Extensive hyperparameter tuning is performed to optimize the Swin Transformer’s performance, considering input size, patch size, embedding size, number of epochs, optimizer type, and regularization technique. Our results demonstrate the Swin Transformer’s superior accuracy, particularly when trained on classically augmented STFT images with optimized settings (320 × 320 input size, RMS optimizer, 8 × 8 patch size, and an embedding size of 128). The approach achieves remarkable testing accuracy (88.37%) and ROC AUC values (94.88%) on the challenging crowdsourced COUGHVID dataset, marking improvements of approximately 2.5% and 11% increases in testing accuracy and ROC AUC values, respectively, compared to previous studies. These findings underscore the efficacy of Swin Transformer architectures in disease detection and healthcare classification problems.

Fallen apple detection as an auxiliary task: Boosting robotic apple detection performance through multi-task learning

In modern agricultural practices, advanced machine learning techniques play a pivotal role in optimizing yields and management. A significant challenge in orchard management is detecting apples on trees, which is essential for effective harvest planning and yield estimation. The YOLO series, especially the YOLOv8 model, stands out as a state-of-the-art solution for object detection, but its potential in orchards remains untapped. Addressing this, our study evaluates YOLOv8’s capability in orchard apple detection, aiming to set a benchmark. By employing image augmentation techniques like exposure, rotation, mosaic, and cutout, we lifted the model's performance to a state-of-the-art level. We further integrated multi-task learning, enhancing tree apple detection by also identifying apples on the ground. This approach resulted in a model with robust accuracy across evaluation metrics. Our results underscore that the YOLOv8 model achieves a leading standard in orchard apple detection. When trained for both tree and fallen apple detection, it outperformed the one when trained exclusively for the former. Recognizing fallen apples not only reduces waste but could also indicate pest activity, influencing strategic orchard decisions and potentially boosting economic returns. Merging cutting-edge tech with agricultural needs, our research showcases the promise of multi-task learning in fruit detection with deep learning.

Diagnosis of glaucoma using multi‐scale attention block in convolution neural network and data augmentation techniques

AbstractGlaucoma is defined as an eye disease leading to vision loss due to the optic nerve damage. It is often asymptomatic, thus, timely diagnosis and treatment is crucial. In this article, we propose a novel approach for diagnosing glaucoma using deep neural networks, trained on fundus images. Our proposed approach involves several key steps, including data sampling, pre‐processing, and classification. To address the data imbalance issue, we employ a combination of suitable image augmentation techniques and Multi‐Scale Attention Block (MAS Block) architecture in our deep neural network model. The MAS Block is a specific architecture design for CNNs that allows multiple convolutional filters of various sizes to capture features at several scales in parallel. This will prevent the over‐fitting problem and increases the detection accuracy. Through extensive experiments with the ACRIMA dataset, we demonstrate that our proposed approach achieves high accuracy in diagnosing glaucoma. Notably, we recorded the highest accuracy (97.18%) among previous studies. The results from this study reveal the potential of our approach to improve early detection of glaucoma and offer more effective treatment strategies for doctors and clinicians in the future. Timely diagnosis plays a crucial role in managing glaucoma since it is often asymptomatic. Our proposed method utilizing deep neural networks shows promise in enhancing diagnostic accuracy and aiding healthcare professionals in making informed decisions.

Étude comparative d'algorithmes d'apprentissage artificiel pour la reconnaissance faciale

Background: The fundamental need for authentication and identification of humans using their physiological, behavioral or biological characteristics, continues to be applied extensively to secure localities, property, financial transactions, etc. Biometric systems based on face characteristics, continue to attract the attention of researchers, major public and private services. In the literature, many methods have been deployed by different authors. The best performance must be found in order to be able to recommend the most effective method. So, the main objective of thisarticle is to make a comparative study of different existing techniques.Methods: A biometric system is generally composed of four stages: acquisition of facial images, preprocessing, extraction of characteristics and finally classification. In this work, the focus is on machine learning algorithms for classification. These algorithms are: Support Vector Machines (SVM), Artificial Neural Networks (ANN), K-Nearest Neighbors (KNN), Random Forests (RF), Logistic Regression (LR), Naive Bayesian Classification (NB: Naive Bayes’ Classifiers) and deep learning techniques such as Convolutional Neural Networks (CNN). The comparison criterion is the average performance, calculated using three performance measures: recognition rate, confusion matrix, and the Area Under Receiver Operating Characteristic (ROC) curve.Results: Based on this criterion, the performance comparison of selected machine learning algorithms, shows that CNN is the best, with an average performance of 100.00% On ORL face database. However, on the YALE database, classical algorithms such as artificial neural networks have obtained the best performances, the highest being a rate of 100%.Discussion: Deep learning techniques are very efficient in image classification as proven by the results on the ORL database. However, their inefficiency on YALE face database is due to the small size of this database which is inappropriate for some deep learning algorithms. But this weakness can be corrected by image augmentation techniques. The comparison of these results with existing state-of-the-art methods is nearly the same. Authors achieved performances of 94.82%, 95.79%, 96.15%, 96.44%, 97.27%, 98.52% and 98.95% for NB, KNN, RF, LR, ANN, SVM and CNN classifiers, respectively. Finally, in depth discussion, it is concluded that between all these approaches which are useful in face recognition, the CNN is the best classification algorithm.

PulmoNet: a novel deep learning based pulmonary diseases detection model

Pulmonary diseases are various pathological conditions that affect respiratory tissues and organs, making the exchange of gas challenging for animals inhaling and exhaling. It varies from gentle and self-limiting such as the common cold and catarrh, to life-threatening ones, such as viral pneumonia (VP), bacterial pneumonia (BP), and tuberculosis, as well as a severe acute respiratory syndrome, such as the coronavirus 2019 (COVID-19). The cost of diagnosis and treatment of pulmonary infections is on the high side, most especially in developing countries, and since radiography images (X-ray and computed tomography (CT) scan images) have proven beneficial in detecting various pulmonary infections, many machine learning (ML) models and image processing procedures have been utilized to identify these infections. The need for timely and accurate detection can be lifesaving, especially during a pandemic. This paper, therefore, suggested a deep convolutional neural network (DCNN) founded image detection model, optimized with image augmentation technique, to detect three (3) different pulmonary diseases (COVID-19, bacterial pneumonia, and viral pneumonia). The dataset containing four (4) different classes (healthy (10,325), COVID-19 (3,749), BP (883), and VP (1,478)) was utilized as training/testing data for the suggested model. The model’s performance indicates high potential in detecting the three (3) classes of pulmonary diseases. The model recorded average detection accuracy of 94%, 95.4%, 99.4%, and 98.30%, and training/detection time of about 60/50 s. This result indicates the proficiency of the suggested approach when likened to the traditional texture descriptors technique of pulmonary disease recognition utilizing X-ray and CT scan images. This study introduces an innovative deep convolutional neural network model to enhance the detection of pulmonary diseases like COVID-19 and pneumonia using radiography. This model, notable for its accuracy and efficiency, promises significant advancements in medical diagnostics, particularly beneficial in developing countries due to its potential to surpass traditional diagnostic methods.

Comparison of simple augmentation transformations for a convolutional neural network classifying medical images

Simple image augmentation techniques, such as reflection, rotation, or translation, might work differently for medical images than they do for regular photographs due to the fundamental properties of medical imaging techniques and the bilateral symmetry of the human body. Here, we compare the predictions of a convolutional neural network (CNN) trained for binary classification by using either no augmentation or one of seven usual types augmentation. We have 11 different medical data sets, mostly related to lung infections or cancer, with X-rays, ultrasound (US) images, and images from positron emission tomography (PET) and magnetic resonance imaging (MRI). According to our results, the augmentation types do not produce statistically significant differences for US and PET data sets, but, for X-rays and MRI images, the best augmentation technique is adding Gaussian blur to images.

A 16 × 16 Patch-Based Deep Learning Model for the Early Prognosis of Monkeypox from Skin Color Images

The DNA virus responsible for monkeypox, transmitted from animals to humans, exhibits two distinct genetic lineages in central and eastern Africa. Beyond the zoonotic transmission involving direct contact with the infected animals’ bodily fluids and blood, the spread of monkeypox can also occur through skin lesions and respiratory secretions among humans. Both monkeypox and chickenpox involve skin lesions and can also be transmitted through respiratory secretions, but they are caused by different viruses. The key difference is that monkeypox is caused by an orthopox-virus, while chickenpox is caused by the varicella-zoster virus. In this study, the utilization of a patch-based vision transformer (ViT) model for the identification of monkeypox and chickenpox disease from human skin color images marks a significant advancement in medical diagnostics. Employing a transfer learning approach, the research investigates the ViT model’s capability to discern subtle patterns which are indicative of monkeypox and chickenpox. The dataset was enriched through carefully selected image augmentation techniques, enhancing the model’s ability to generalize across diverse scenarios. During the evaluation phase, the patch-based ViT model demonstrated substantial proficiency, achieving an accuracy, precision, recall, and F1 rating of 93%. This positive outcome underscores the practicality of employing sophisticated deep learning architectures, specifically vision transformers, in the realm of medical image analysis. Through the integration of transfer learning and image augmentation, not only is the model’s responsiveness to monkeypox- and chickenpox-related features enhanced, but concerns regarding data scarcity are also effectively addressed. The model outperformed the state-of-the-art studies and the CNN-based pre-trained models in terms of accuracy.

An Exploration of Deep Learning Techniques for the Detection of Grape Diseases

Abstract: Plant diseases are one of the major contributors to economic loss in the agriculture industry worldwide. Detection of disease at early stages can help in the reduction of this loss. In recent times, a lot of emphasis has been done on disease detection due to the overall increase in production as well as the loss of grape number. With deep learning, having a promising future and having the advantages of automatic learning and feature extraction, the use of these techniques has now been widely spread. This paper reviewed the existing deep-learning techniques available for grape disease detection. Firstly, covering the various steps in a grape disease detection model ranging from the various sources of image acquisition, the different image augmentation techniques and the various models used, and the parameters required to evaluate. Secondly, the study summarizes the important findings of all literature available on the theme. The paper also tries to highlight the various challenges faced by the researchers and the common trend among them, so that future research on the topic can achieve higher performance.

Udderly accurate: A deep learning based modelling for determination of dairyness of Sahiwal cow using computer vision

Dairy farming is a crucial agricultural practice for food and nutritional security. Selection of the best milching animals has always been a challenging task for optimising production efficiency. The efficacy of milk yield production is contingent upon a number of factors including inherent linear traits. Linear traits refer to the quantifiable physical characteristics that are associated with the production and reproduction capabilities of dairy animals. This article presents an innovative approach for the classification of Sahiwal cows into high, medium and low yielder categories based on images featuring linear traits through emerging deep learning and computer vision techniques. A large dataset of 4110 images highlighting important linear traits such as udder size, shape, and texture of different categories of Sahiwal cows has been created for training, validation and testing of the model. Images were collected from the herd of Sahiwal cows maintained at the National Dairy Research Institute, Karnal. The dataset was pre-processed using image augmentation techniques to enhance the model’s robustness. Different architectures of CNN models namely InceptionV3, ResNet50 and GoogleNet were trained and optimised. The Inception V3 model demonstrated the best result with 85.64% testing accuracy among all these models in classifying the cow. The developed model can be used under field conditions to determine the dairyness of a cow in real time mode in place of human experts. Additionally, the model’s interpretability is evaluated through feature visualisation, showcasing the importance of different udder features in milk yield prediction.

Fatigue damage reconstruction in glass/epoxy composites via thermal analysis and machine learning: A theoretical study

This study introduces an advanced, non-contact diagnostic tool for structural health monitoring of fatigue damage in fiber/polymer composite materials. The approach combines thermal image recognition of fatigue self-heating hotspots with high-fidelity thermal modeling to quantitatively assess subsurface fatigue damage distributions by machine learning. To this end, artificial thermal images are generated through 3D numerical thermal analysis of an inherent fatigue damage heat source within a glass/epoxy composite, derived from sampling a multivariate Gaussian distribution of microcracks. Subsequently, these synthetic thermal images are employed to train three distinct regression models: a convolutional neural network, a Gaussian processes regressor, and a straightforward least squares model. Various image augmentation techniques are applied to expand the dataset efficiently. All models accurately predict the size of the damage and – most importantly – the maximum temperature within the damage deep inside the composite. The regression methods estimate the diagonal elements of covariance matrix components of the Gaussian distribution, with accuracies ranging from 86% to 99%. The findings presented in this work contribute to establishing a solid foundation for non-destructive subsurface fatigue damage assessment in composite materials, with many practical applications in experimental composites fatigue research.