General Convolutional Neural Network Research Articles

Remote sensing-enhanced transfer learning approach for agricultural damage and change detection: A deep learning perspective

With the continuous advancement of science and technology, there has been a growing awareness of safety among people worldwide. Natural disasters such as wildfires, earthquakes, and floods pose persistent threats to both lives and property on our planet, which serves as our fundamental habitat. While it is impossible to prevent or entirely avert these calamities, rapid identification of affected areas and prompt damage assessment post-disaster can significantly aid in the formulation of effective rescue strategies, ultimately saving more lives. This article delves into the application of transfer learning in satellite image damage assessment—a methodology that involves transferring previously acquired knowledge to enhance a model's adaptability to new tasks. Given the limited availability of datasets for satellite image analysis, transfer learning proves to be an effective approach. Specifically, the study proposes a transfer learning method based on YOLOv5 for satellite image damage assessment. Initially, a general convolutional neural network model is trained using a substantial dataset of natural images. Subsequently, the early layers of this model are frozen, while the later layers undergo training to adapt to satellite image data. Fine-tuning is then employed to further enhance the overall model performance. The results demonstrate that this approach yields a high accuracy rate in satellite image damage assessment. Moreover, compared to conventional deep learning methods, the proposed method effectively leverages pre-trained models' knowledge, thereby reducing data dependency. Additionally, it displays robust generalization capabilities across diverse tasks and datasets, underscoring its potential for facilitating transfer learning across various domains.

Game-theoretic policy computing and simulation for blockchained buffering system via diffusion approximation

Abstract We study 2-stage game-theoretic problem oriented 3-stage service policy computing, convolutional neural network (CNN) based algorithm design, and simulation for a blockchained buffering system with federated learning. More precisely, based on the game-theoretic problem consisting of both “win-lose” and “win-win” 2-stage competitions, we derive a 3-stage dynamical service policy via a saddle point to a zero-sum game problem and a Nash equilibrium point to a non-zero-sum game problem. This policy is concerning users-selection, dynamic pricing, and online rate resource allocation via stable digital currency for the system. The main focus is on the design and analysis of the joint 3-stage service policy for given queue/environment state dependent pricing and utility functions. The asymptotic optimality and fairness of this dynamic service policy is justified by diffusion modeling with approximation theory. A general CNN based policy computing algorithm flow chart along the line of the so-called big model framework is presented. Simulation case studies are conducted for the system with three users, where only two of the three users can be selected into the service by a zero-sum dual cost game competition policy at a time point. Then, the selected two users get into service and share the system rate service resource through a non-zero-sum dual cost game competition policy. Applications of our policy in the future blockchain based Internet (e.g., metaverse and web3.0) and supply chain finance are also briefly illustrated.

DSTCNet: Deep Spectro-Temporal-Channel Attention Network for Speech Emotion Recognition.

Speech emotion recognition (SER) plays an important role in human-computer interaction, which can provide better interactivity to enhance user experiences. Existing approaches tend to directly apply deep learning networks to distinguish emotions. Among them, the convolutional neural network (CNN) is the most commonly used method to learn emotional representations from spectrograms. However, CNN does not explicitly model features' associations in the spectral-, temporal-, and channel-wise axes or their relative relevance, which will limit the representation learning. In this article, we propose a deep spectro-temporal-channel network (DSTCNet) to improve the representational ability for speech emotion. The proposed DSTCNet integrates several spectro-temporal-channel (STC) attention modules into a general CNN. Specifically, we propose the STC module that infers a 3-D attention map along the dimensions of time, frequency, and channel. The STC attention can focus more on the regions of crucial time frames, frequency ranges, and feature channels. Finally, experiments were conducted on the Berlin emotional database (EmoDB) and interactive emotional dyadic motion capture (IEMOCAP) databases. The results reveal that our DSTCNet can outperform the traditional CNN-based and several state-of-the-art methods.

Optimization of Briquette Classification Using Deep Learning

Indonesia needs energy to meet its needs. Energy sourced from waste is called biomass briquettes. The manufacture of briquettes is still done in a traditional way, so there are product quality problems. A computational approach has been taken to base the quality of the product on certain characteristics so that the types of briquettes that are made can be classified or sorted. The research objective of this work is to determine the quality of briquettes. The approach uses deep learning methods and convolutional neural network (CNN) architecture. Classification is based on good and bad briquette products, and testing is based on the level of performance accuracy. The dataset formed consists of 5,280 images. As training data, 85% of data is used and 15% is used as test data using 300 epoch parameters, 32 batch sizes, and learning speed up to 0.001. The results of testing and evaluating the performance of briquette products with the general CNN architecture have a level of accuracy that is not optimal. The optimization results of testing the MobileNetV2 architecture with a ratio of 70:30 obtains predictions with 0.99 as the highest accuracy value and 0.73 as the lowest. The average predicted value of the model is 0.95. The computational approach is able to provide traditional communities with solutions for the process of making briquettes.

Performance evaluation of deep learning algorithms for heat loss damage classification in buildings from UAV-borne infrared images

Deep learning (DL) algorithms are gaining popularity to automate anomaly detection in a variety of civil and mechanical engineering domains. Nevertheless, very few attempts have been made to use infrared (IR) images with DL algorithms to detect heat loss in a building and expedite the otherwise cumbersome energy audit process. However, no clear guidelines exist for the selection of appropriate algorithms or hyperparameters (e.g., number of epochs, learning rate, number of images) to implement DL with IR images. To address the research gap, a comparative sensitivity analysis of four well-known DL algorithms for the classification of heat losses using only IR images was performed for the first time. Four DL algorithms, namely, general convolutional neural network (CNN), VGG16, transfer learning (TL) VGG16, and TL inceptionV3 were selected and used for automating the detection of heat loss from IR images. The sensitivity of the selected CNN algorithms was analyzed based on i) the size of the IR images dataset, ii) the number of epochs, and iii) the effect of an unbalanced dataset. The goal was to provide a framework that may aid in the selection of suitable CNN architecture as well as parameters related to data size and an optimal number of training iterations for reliable heat loss identification using IR images. This article has provided a framework for automated heat loss detection that can be applied to any IR dataset without the need to pre-process the IR images. The accuracy of the selected DL algorithms is validated with IR images of laboratory setups and real-world buildings characterized by different types of defects. The comparative sensitivity analysis was conducted based on multiple performance indicators such as precision, recall, and F-1 score which resulted in optimal values of hyperparameters and the most suitable DL algorithm. The automated identification of heat loss yielded performance indicator values of more than 90%.

Beyond REM: A New Approach to the Use of Image Classifiers for the Management of 6G Networks.

The management of cellular networks, particularly within the environment rapidly advancing to 6G, presents considerable challenges due to the highly dynamic radio environment. Traditional tools such as Radio Environment Maps (REMs) have proven inadequate for real-time network changes, underlining the need for more sophisticated solutions. In response to these challenges, this work introduces a novel approach that harnesses the unprecedented power of state-of-the-art image classifiers for network management. This method involves the generation of Network Synthetic Images (NSIs), which are enriched heat maps that precisely reflect varying cellular network operating states. Created from user location traces linked with Key Performance Indicators (KPIs), NSIs are strategically designed to meet the intricate demands of 6G networks. This research delves deep into a comprehensive analysis of the diverse factors that could potentially impact the successful application of this methodology in the realm of 6G. The results from this investigation, coupled with a comparative assessment against traditional REM usage, emphasize the superior performance of this innovative method. Additionally, a case study involving an automatic network diagnosis scenario validates the effectiveness of this approach. The findings reveal that a generic Convolutional Neural Network (CNN), one of the most powerful tools in the arsenal of modern image classifiers, delivers enhanced performance, even with a reduced demand for positioning accuracy. This contributes significantly to the real-time, robust management of cellular networks as we transition into the era of 6G.

GuidedNet: A General CNN Fusion Framework via High-Resolution Guidance for Hyperspectral Image Super-Resolution

Hyperspectral image super-resolution (HISR) is about fusing a low-resolution hyperspectral image (LR-HSI) and a high-resolution multispectral image (HR-MSI) to generate a high-resolution hyperspectral image (HR-HSI). Recently, convolutional neural network (CNN)-based techniques have been extensively investigated for HISR yielding competitive outcomes. However, existing CNN-based methods often require a huge amount of network parameters leading to a heavy computational burden, thus, limiting the generalization ability. In this article, we fully consider the characteristic of the HISR, proposing a general CNN fusion framework with high-resolution guidance, called GuidedNet. This framework consists of two branches, including 1) the high-resolution guidance branch (HGB) that can decompose the high-resolution guidance image into several scales and 2) the feature reconstruction branch (FRB) that takes the low-resolution image and the multiscaled high-resolution guidance images from the HGB to reconstruct the high-resolution fused image. GuidedNet can effectively predict the high-resolution residual details that are added to the upsampled HSI to simultaneously improve spatial quality and preserve spectral information. The proposed framework is implemented using recursive and progressive strategies, which can promote high performance with a significant network parameter reduction, even ensuring network stability by supervising several intermediate outputs. Additionally, the proposed approach is also suitable for other resolution enhancement tasks, such as remote sensing pansharpening and single-image super-resolution (SISR). Extensive experiments on simulated and real datasets demonstrate that the proposed framework generates state-of-the-art outcomes for several applications (i.e., HISR, pansharpening, and SISR). Finally, an ablation study and more discussions assessing, for example, the network generalization, the low computational cost, and the fewer network parameters, are provided to the readers. The code link is: https://github.com/Evangelion09/GuidedNet.

From Lab to Real World: Assessing the Effectiveness of Human Activity Recognition and Optimization through Personalization.

Human activity recognition (HAR) algorithms today are designed and evaluated on data collected in controlled settings, providing limited insights into their performance in real-world situations with noisy and missing sensor data and natural human activities. We present a real-world HAR open dataset compiled from a wristband equipped with a triaxial accelerometer. During data collection, participants had autonomy in their daily life activities, and the process remained unobserved and uncontrolled. A general convolutional neural network model was trained on this dataset, achieving a mean balanced accuracy (MBA) of 80%. Personalizing the general model through transfer learning can yield comparable and even superior results using fewer data, with the MBA improving to 85%. To emphasize the issue of insufficient real-world training data, we conducted training of the model using the public MHEALTH dataset, resulting in 100% MBA. However, upon evaluating the MHEALTH-trained model on our real-world dataset, the MBA drops to 62%. After personalizing the model with real-world data, an improvement of 17% in the MBA is achieved. This paper showcases the potential of transfer learning to make HAR models trained in different contexts (lab vs. real-world) and on different participants perform well for new individuals with limited real-world labeled data available.

A General Convolutional Neural Network to Reconstruct Remotely Sensed Chlorophyll-a Concentration

Satellite-observed chlorophyll-a (Chl-a) concentrations are key to studies of phytoplankton dynamics. However, there are gaps in remotely sensed images mainly due to cloud coverage which requires reconstruction. This study proposed a method to build a general convolutional neural network (CNN) model that can reconstruct images in unfamiliar areas. Although several CNN models to reconstruct Chl-a in a specific area have already been proposed, the model in this research has the advantage of generality. The model uses a more flexible U-net architecture so that it can accept input of different shapes. Images from three areas of different shapes were used in model training to improve the generality of the model. Six models, with different auxiliary input schemes and architectures, were trained and evaluated. Results show that the model with bathymetry input and coarse-to-fine architecture has the best performance and can give reasonable reconstruction for the unfamiliar area. The best model shows better results than traditional interpolation methods when reconstructing for an unfamiliar area, especially in regions outside the data coverage.

Double Branch Parallel Network for Segmentation of Buildings and Waters in Remote Sensing Images

The segmentation algorithm for buildings and waters is extremely important for the efficient planning and utilization of land resources. The temporal and space range of remote sensing pictures is growing. Due to the generic convolutional neural network’s (CNN) insensitivity to the spatial position information in remote sensing images, certain location and edge details can be lost, leading to a low level of segmentation accuracy. This research suggests a double-branch parallel interactive network to address these issues, fully using the interactivity of global information in a Swin Transformer network, and integrating CNN to capture deeper information. Then, by building a cross-scale multi-level fusion module, the model can combine features gathered using convolutional neural networks with features derived using Swin Transformer, successfully extracting the semantic information of spatial information and context. Then, an up-sampling module for multi-scale fusion is suggested. It employs the output high-level feature information to direct the low-level feature information and recover the high-resolution pixel-level features. According to experimental results, the proposed networks maximizes the benefits of the two models and increases the precision of semantic segmentation of buildings and waters.

Classifying Brain Tumors on Magnetic Resonance Imaging by Using Convolutional Neural Networks

The study of neuroimaging is a very important tool in the diagnosis of central nervous system tumors. This paper presents the evaluation of seven deep convolutional neural network (CNN) models for the task of brain tumor classification. A generic CNN model is implemented and six pre-trained CNN models are studied. For this proposal, the dataset utilized in this paper is Msoud, which includes Fighshare, SARTAJ, and Br35H datasets, containing 7023 MRI images. The magnetic resonance imaging (MRI) in the dataset belongs to four classes, three brain tumors, including Glioma, Meningioma, and Pituitary, and one class of healthy brains. The models are trained with input MRI images with several preprocessing strategies applied in this paper. The CNN models evaluated are Generic CNN, ResNet50, InceptionV3, InceptionResNetV2, Xception, MobileNetV2, and EfficientNetB0. In the comparison of all CNN models, including a generic CNN and six pre-trained models, the best CNN model for this dataset was InceptionV3, which obtained an average Accuracy of 97.12%. The development of these techniques could help clinicians specializing in the early detection of brain tumors.

Integrated CNN and Waterwheel Plant Algorithm for Enhanced Global Traffic Detection

Traffic detection is critical in ensuring road safety and efficient traffic management, demanding deploying accurate and practical algorithms. This research explores the fusion of Convolutional Neural Networks (CNNs) and the Waterwheel Plant Algorithm to augment global traffic detection capabilities, utilizing a diverse dataset primarily collected from Turkey. A comprehensive evaluation of prominent CNN architectures, such as VGG19Net, AlexNet, ResNet-50, GoogLeNet, and a generic CNN, underscores substantial efficacy, with the CNN achieving an accuracy of 92.14%. Introducing the Waterwheel Plant Algorithm (WWPA) further enhances performance, as exemplified by the hybrid WWPA-CNN model, exhibiting an impressive accuracy of 97.28%. These findings highlight the promising synergies between traditional optimization algorithms and advanced neural networks, showcasing the potential for innovative developments in traffic monitoring systems and broader applications within computer vision. The statistical analyses, encompassing ANOVA and the Wilcoxon Signed Rank Test, robustly underscore the significance of this integrated approach. As the research contributes to the evolution of traffic monitoring systems, these insights provide a solid foundation for advancements in the field, fostering innovation and shaping the future landscape of computer vision applications.

A survey of fine-grained visual categorization based on deep learning

Deep learning has achieved excellent results in various tasks in the field of computer vision, especially in fine-grained visual categorization. It aims to distinguish the subordinate categories of the label-level categories. Due to high intra-class variances and high inter-class similarity, the fine-grained visual categorization is extremely challenging. This paper first briefly introduces and analyzes the related public datasets. After that, some of the latest methods are reviewed. Based on the feature types, the feature processing methods, and the overall structure used in the model, we divide them into three types of methods: methods based on general convolutional neural network (CNN) and strong supervision of parts, methods based on single feature processing, and methods based on multiple feature processing. Most methods of the first type have a relatively simple structure, which is the result of the initial research. The methods of the other two types include models that have special structures and training processes, which are helpful to obtain discriminative features. We conduct a specific analysis on several methods with high accuracy on public datasets. In addition, we support that the focus of the future research is to solve the demand of existing methods for the large amount of the data and the computing power. In terms of technology, the extraction of the subtle feature information with the burgeoning vision transformer (ViT) network is also an important research direction.

Novel CNN Accelerator Design With Dual Benes Network Architecture

We presented a novel hardware architecture that uses dual Benes networks to accelerate Convolutional Neural Network (CNN) algorithms. This architecture can reduce the need for high-speed buses and maintain a high-speed connection between execution units and memories. Also, this design can work with multiple neural network models by changing the Benes network configuration only due to the reorderability of the non-blocking networks. The proposed architecture can save the time of re-implementing the hardware design. Thus, this architecture can act as a general CNN accelerator. We successfully implemented our CNN accelerator on the Xilinx Virtex UltraScale+ VU19P and evaluated it by running the CNN model LeNet-5.

Analysis of convolutional neural network image classifiers in a hierarchical max-pooling model with additional local pooling

In practical applications of image classification, methods based on convolutional neural networks are used as standard nowadays. However, the success of these methods is not explained sufficiently from a theoretical point of view. For this, a statistical model for image classification, namely a hierarchical max-pooling model with additional local pooling, is introduced. Here the additional local pooling enables the hierarchical model to combine parts of the image which have a variable relative distance towards each other. Various convolutional neural network image classifiers are introduced and analyzed in view of their rate of convergence of the misclassification risk of the estimates towards the optimal misclassification risk. This analysis provides a theoretical explanation why general convolutional neural network architectures that include some kind of local pooling layers are useful in some image classification situations and gives theoretical hints for choosing the right network architecture. Furthermore, the finite sample size performance of these network architectures is illustrated by applying them to simulated and real data.

Adjacent age classification algorithm of yellow-feathered chickens based on multi-scale feature fusion

The age of yellow-feathered chicken is important to distinguish the freshness of meat quality in the trade of yellow-feathered chicken. To investigate whether the Convolutional Neural Network (CNN) model can be applied to the instar classification of yellow-feathered chickens, a multi-scale feature fusion model called Chicken_Age_Network (CANet) was proposed. The model uses Inception to construct a feature extraction layer and extract the feature information of each chicken face image to improve the classification accuracy of the model. First, the self-developed yellow-feathered chicken facial image collection application was used to collect images, and the yellow-feathered chicken image database was constructed by using day-age classification. Second, the standard face image of yellow-feathered chicken was obtained by using Structural Similarity Index Method detection(SSIM), image segmentation, background removal, and normalization. The adjacent age classification needs to extract more features, and CANet’s feature extraction layer based on multi-scale feature fusion can extract features of different sizes. Another advantage of CANet is that the GMP (Global Max Pooling) replaces the final fully connected layer of the general CNN to reduce parameters and optimize the network model. Finally, chicken face images of adjacent days of age were tested and compared on VGG13, VGG19, DenseNet121, DenseNet161, SE-ResNet-20, MobileNet V1, ShuffleNet G2, ResNet50, ResNet34 and CSPDenseNet121. Test results show that CANet can quickly and accurately identify the age of yellow-feathered chickens. The classification accuracy on CANet model is 96.29 %, which is better than VGG13 (93.09 %), VGG19 (95.08 %), ResNet50 (85.53 %), ResNet34 (91.14 %), DenseNet121 (93.03 %), DenseNet161 (93.33 %), MobileNet V1(86.06 %), ShuffleNet G2(93.18 %), CSPDenseNet121(84.76 %) and SE-ResNet-20 (83.86 %). In order to verify the generalization of the model, experiments were carried out on the public data set CIFAR-10. Results show that CANet has good generalization and can be applied to other classification problems.

Modified layer deep convolution neural network for text-independent speaker recognition

peaker recognition is the task of identifying the spokesman automatically using speaker-specific features. It has been a popular and most involved topic in the field of speech technology. This field opens a wide opportunity for research and finds its application in the areas such as forensics, authentication, security, etc. In this work, a modified deep-convolutional neural network structure has been proposed for speaker identification that has improved convolution, activation, and pooling layers along with Adam’s optimiser. The proposed architecture yielded the increase of prediction accuracy and reduction of Loss function when compared to the generic Convolutional Neural Network scheme. The execution of the proposed architecture is validated by various datasets and the outcomes show that the modified CNN performs better than the other state-of-the-art models regarding both accuracy (avg 99%) and loss function (avg 1%). From the analysis, it is found that the Modified-CNN suits the best for real-time speaker identification applications as the efficacy of the model does not degrade due to the effects of noise and interferences that are caused in the recording environment. Relevance of the work: Speaker Recognition is an area of interest in which ML and DL schemes, when combined, have the potential to make history in the areas of Automation and Authentication. Using a modified CNN can enhance the process by ignoring many issues such as false positives, background noise, and so on. This process can be expanded to create a Raga Identification and Therapy mechanism that can be used to treat diseases.

Divide-and-Attention Network for HE-Stained Pathological Image Classification.

Simple SummaryWe propose a Divide-and-Attention network that can learn representative pathological image features with respect to different tissue structures and adaptively focus on the most important ones. In addition, we introduce deep canonical correlation analysis constraints in the feature fusion process of different branches, so as to maximize the correlation of different branches and ensure that the fused branches emphasize specific tissue structures. Extensive experiments on three different pathological image datasets show that the proposed method achieved competitive results.Since pathological images have some distinct characteristics that are different from natural images, the direct application of a general convolutional neural network cannot achieve good classification performance, especially for fine-grained classification problems (such as pathological image grading). Inspired by the clinical experience that decomposing a pathological image into different components is beneficial for diagnosis, in this paper, we propose a Divide-and-Attention Network (DANet) for Hematoxylin-and-Eosin (HE)-stained pathological image classification. The DANet utilizes a deep-learning method to decompose a pathological image into nuclei and non-nuclei parts. With such decomposed pathological images, the DANet first performs feature learning independently in each branch, and then focuses on the most important feature representation through the branch selection attention module. In this way, the DANet can learn representative features with respect to different tissue structures and adaptively focus on the most important ones, thereby improving classification performance. In addition, we introduce deep canonical correlation analysis (DCCA) constraints in the feature fusion process of different branches. The DCCA constraints play the role of branch fusion attention, so as to maximize the correlation of different branches and ensure that the fused branches emphasize specific tissue structures. The experimental results of three datasets demonstrate the superiority of the DANet, with an average classification accuracy of 92.5% on breast cancer classification, 95.33% on colorectal cancer grading, and 91.6% on breast cancer grading tasks.

Convolutional neural network based simulation and analysis for backward stochastic partial differential equations

We develop a generic convolutional neural network (CNN) based numerical scheme to simulate the 2-tuple adapted strong solution to a unified system of backward stochastic partial differential equations (B-SPDEs) driven by Brownian motions, which can be used to model many real-world system dynamics such as optimal control and differential game problems. The dynamics of the scheme is modeled by a CNN through conditional expectation projection. It consists of two convolution parts: W layers of backward networks and L layers of reinforcement iterations. Furthermore, it is a completely discrete and iterative algorithm in terms of both time and space with mean-square error estimation and almost sure (a.s.) convergence supported by both theoretical proof and numerical examples. In doing so, we need to prove the unique existence of the 2-tuple adapted strong solution to the system under both conventional and Malliavin derivatives with general local Lipschitz and linear growth conditions.

Pseudoprospective Paraclinical Interaction of Radiology Residents With a Deep Learning System for Prostate Cancer Detection: Experience, Performance, and Identification of the Need for Intermittent Recalibration.

The aim of this study was to estimate the prospective utility of a previously retrospectively validated convolutional neural network (CNN) for prostate cancer (PC) detection on prostate magnetic resonance imaging (MRI). The biparametric (T2-weighted and diffusion-weighted) portion of clinical multiparametric prostate MRI from consecutive men included between November 2019 and September 2020 was fully automatically and individually analyzed by a CNN briefly after image acquisition (pseudoprospective design). Radiology residents performed 2 research Prostate Imaging Reporting and Data System (PI-RADS) assessments of the multiparametric dataset independent from clinical reporting (paraclinical design) before and after review of the CNN results and completed a survey. Presence of clinically significant PC was determined by the presence of an International Society of Urological Pathology grade 2 or higher PC on combined targeted and extended systematic transperineal MRI/transrectal ultrasound fusion biopsy. Sensitivities and specificities on a patient and prostate sextant basis were compared using the McNemar test and compared with the receiver operating characteristic (ROC) curve of CNN. Survey results were summarized as absolute counts and percentages. A total of 201 men were included. The CNN achieved an ROC area under the curve of 0.77 on a patient basis. Using PI-RADS ≥3-emulating probability threshold (c3), CNN had a patient-based sensitivity of 81.8% and specificity of 54.8%, not statistically different from the current clinical routine PI-RADS ≥4 assessment at 90.9% and 54.8%, respectively ( P = 0.30/ P = 1.0). In general, residents achieved similar sensitivity and specificity before and after CNN review. On a prostate sextant basis, clinical assessment possessed the highest ROC area under the curve of 0.82, higher than CNN (AUC = 0.76, P = 0.21) and significantly higher than resident performance before and after CNN review (AUC = 0.76 / 0.76, P ≤ 0.03). The resident survey indicated CNN to be helpful and clinically useful. Pseudoprospective paraclinical integration of fully automated CNN-based detection of suspicious lesions on prostate multiparametric MRI was demonstrated and showed good acceptance among residents, whereas no significant improvement in resident performance was found. General CNN performance was preserved despite an observed shift in CNN calibration, identifying the requirement for continuous quality control and recalibration.