Average Process Research Articles

License Plate Detection and Recognition in Unconstrained Environment Using Deep Learning

Real-time detection and recognition systems for vehicle license plates present a significant design and implementation challenge, arising from factors such as low image resolution, data noise, and various weather and lighting conditions. This study presents an efficient automated system for the identification and classification of vehicle license plates, utilizing deep learning techniques. The system is specifically designed for Iraqi vehicle license plates, adapting to various backgrounds, different font sizes, and non-standard formats. The proposed system has been designed to be integrated into an automated entrance gate security system. The system's framework encompasses two primary phases: license plate detection (LPD) and character recognition (CR). The utilization of the advanced deep learning technique YOLOv4 has been implemented for both phases owing to its adeptness in real-time data processing and its remarkable precision in identifying diminutive entities like characters on license plates. In the LPD phase, the focal point is on the identification and isolation of license plates from images, whereas the CR phase is dedicated to the identification and extraction of characters from the identified license plates. A substantial dataset comprising Iraqi vehicle images captured under various lighting and weather circumstances has been amassed for the intention of both training and testing. The system attained a noteworthy accuracy level of 95.07\%, coupled with an average processing time of 118.63 milliseconds for complete end-to-end operations on a specified dataset, thus highlighting its suitability for real-time applications. The results suggest that the proposed system has the capability to significantly enhance the efficiency and reliability of vehicle license plate recognition in various environmental conditions, thus making it suitable for implementation in security and traffic management contexts.

Analysis of Intensive Longitudinal Data: Putting Psychological Processes in Perspective.

Research based on intensive longitudinal data (ILD)-consisting of many repeated measures from one or multiple individuals-is rapidly gaining popularity in psychological science. To appreciate the unique potential of ILD research for clinical psychology, this review begins by examining how our three traditional research approaches fall short when the goal is to investigate processes. It then explores how the analysis of ILD can be used to study a process as it unfolds within a specific person over time but also to study average process features or individual differences therein. By emphasizing the alignment between research questions, data collection, and analytical strategies, the potential of ILD research is further highlighted. It is argued that for future progress it is essential to integrate machine learning and causal inference methods with statistical techniques for ILD and to become more explicit about timescales, time frames, and dynamics in psychological theories.

IncARMAG: A convolutional neural network with multi-level autoregressive moving average graph convolutional processing framework for medical image classification

Effective computer-aided detection and diagnosis algorithms are being sought to carry out complex pattern recognition with great precision while increasing efficiency and robustness through automation. In this study, we leveraged self-constructing graph (SCG) and autoregressive moving average (ARMA) graph convolution to postprocess multi-level feature maps obtained from an Inception V3 network for medical image classification. The adopted SCG learns latent embeddings from the feature maps, which are subsequently used to construct graph topology based on embedding similarity. The ARMA convolution then learns to encode more flexible spectral filter responses that simultaneously takes into account long-range dependencies from the SCG graph. The proposed IncARMAG model was evaluated on 2D MedMNIST datasets, which consist of several imaging modalities and classification tasks in the medical domain. Results from our experiments on MedMNIST datasets showed that IncARMAG outperformed state-of-the-art models on many cases. IncARMAG has also achieved the highest accuracy based on evaluation of various benchmark CNN, transformers, and hybrid models on four selected datasets. These results highlight how IncARMAG is a strong candidate in medical image analysis involving classification tasks.

Predicted missing information biases ensemble perception of temporally ordered facial expressions

By observing dynamically changing facial expressions, humans can use a specialized capacity known as ensemble coding to effortlessly obtain a summary representation of an individual’s emotional state. However, few studies have examined whether the missing expression informed by the statistical regularity in the changing facial expressions can be sampled and then influence the perceptual averaging process. In Experiment 1a and 1b, we manipulated the amount of prior information from local regularity by varying the position of the missing expression in the temporal sequence (1a: Neutral to Disgust and/or Disgust to Neutral,1b: Neutral to Happy and/or Happy to Neutral) within a trial. Results showed that ensemble estimates were towards the mean of expressions including both the presented and the missing faces, only when sufficient predictability (e.g. a missing expression in the late position) informed by local regularity. In Experiment 2, we added prior information from global regularities to help boost the predictability of an early missing expression by keeping the emotional direction consistent in a block. However, estimates were not towards the mean of expressions including both the presented and the missing expressions as expected. Although the generalizability may be limited, these findings suggest that prior information at different levels of hierarchical predictive coding may exert qualitatively different influences on the perceptual averaging of temporally ordered facial expressions with missing items.

Improving Efficiency: Optimizing the Average Processing Time of Crm Systems Based on Machine Learning

Improving Efficiency: Optimizing the Average Processing Time of Crm Systems Based on Machine Learning

Automated Video Assistant Referee in Lead Climbing

Introduction In lead climbing, points are awarded for each hold held, which are then primarily decisive for the ranking. These points are awarded immediately after an attempt by a judge who observed the athlete. Competition organisers are required to make video recordings, but these are only viewed by the head referee in case of an appeal. Our work aims to enhance the objectivity of judging. Our approach is based on applying pose estimators to video footage that is recorded anyway, and a subsequent analysis that determines metrics for judging. We want to tune our approach based on the scores given by judges while considering regulations of the International Federation of Sport Climbing. Methods Our video footage was filmed by the Swiss Alpine Club for judging at the European Climbing Championships 2024 in Villar and at the Swiss Championships 2024 in Lenzburg. Initially, twenty videos (30 seconds each including the fall from the wall towards the end) were extracted. Half of the videos were cropped to focus on the athlete. Thus, a total of ten videos from Villar (outdoor) were analysed – five uncropped (five men), five with focus (women) – and ten videos from Lenzburg (indoor) – five uncropped (five women), five with focus (five men). As a first step, two pose estimators were evaluated regarding processing speed in frames per second (FPS) and percentage of frames in which a pose was detected: BlazePose from Google which has already been used for climbing research (Kim et al., 2023) and Real-Time Multi-Person Pose Estimation (RTMO) (Lu et al., 2024). Results BlazePose demonstrated a higher average processing speed compared to RTMO: 22 FPS vs. 2 FPS. In uncropped videos, BlazePose detected a pose in 5% of outdoor video footage and 33% of indoor video footage. In close-up videos, BlazePose detected poses in 99% of outdoor video footage and 93% of indoor video footage. RTMO achieved a 100% pose detection rate across all videos whereby a subsequent inspection showed that many false poses were considered as “detected”. Discussion/Conclusion An automated video assistant referee should decide quickly. RTMO seems too slow with the given hardware (AMD Ryzen 7, 3.2GHz, 16GB RAM), since pose estimation of a 5s video sequence (150 frames) would take more than a minute. The video would also have to be cropped before each pose estimation, otherwise few (BlazePose) or incorrect (RTMO) poses would be detected caused by picture content being more contrasting and larger than the athlete. Next, we will compare our scoring algorithm representing given regulations with scores awarded. The results will then be discussed at the next educational event for judges at national and international level.

Estimation of stationary and non-stationary moving average processes in the correlation domain.

This paper introduces a novel approach for the offline estimation of stationary moving average processes, further extending it to efficient online estimation of non-stationary processes. The novelty lies in a unique technique to solve the autocorrelation function matching problem leveraging that the autocorrelation function of a colored noise is equal to the autocorrelation function of the coefficients of the moving average process. This enables the derivation of a system of nonlinear equations to be solved for estimating the model parameters. Unlike conventional methods, this approach uses the Newton-Raphson and Levenberg-Marquardt algorithms to efficiently find the solution. A key finding is the demonstration of multiple symmetrical solutions and the provision of necessary conditions for solution feasibility. In the non-stationary case, the estimation complexity is notably reduced, resulting in a triangular system of linear equations solvable by backward substitution. For online parameter estimation of non-stationary processes, a new recursive formula is introduced to update the sample autocorrelation function, integrating exponential forgetting of older samples to enable parameter adaptation. Numerical experiments confirm the method's effectiveness and evaluate its performance compared to existing techniques.

Image Similarity Quantum Algorithm and Its Application in Image Retrieval Systems

The measurement of image similarity represents a fundamental task within the domain of image processing, enabling the application of sophisticated computational techniques to ascertain the degree of similarity between two images. To enhance the performance of these similarity measurement algorithms, the academic community has investigated a range of quantum algorithms. Notably, the swap test-based quantum inner product algorithm (ST-QIP) has emerged as a pivotal method for computing image similarity. However, the inherent destructive nature of the swap test necessitates multiple quantum state evolutions and measurements, which leads to consumption of quantum resources and prolonged computational time, ultimately constraining its practical applicability. To address these limitations, this study introduces an advanced quantum inner product algorithm based on amplitude estimation (AE-QIP) designed to compute image similarity. This innovative approach circumvents the repetitive measurement processes associated with the swap test, thereby optimizing the utilization of quantum resources and substantially enhancing the algorithm’s performance. We conducted experiments using a quantum simulator to implement the AE-QIP algorithm and evaluate its effectiveness in the image retrieval tasks. It is found that the AE-QIP algorithm achieves comparable precision to the ST-QIP algorithm while exhibiting significant reductions in qubit consumption and average processing time. Additionally, our findings suggest that increasing the number of ancillary qubits can further enhance the accuracy of the AE-QIP algorithm. Overall, within the acceptable error thresholds, the AE-QIP algorithm exhibits enhanced efficiency relative to the ST-QIP algorithm. However, significant further research is needed to address the challenges involved in optimizing the performance of quantum retrieval systems as a whole.

A Novel Approach for the Counting of Wood Logs Using cGANs and Image Processing Techniques

This study tackles the challenge of precise wood log counting, where applications of the proposed methodology can span from automated approaches for materials management, surveillance, and safety science to wood traffic monitoring, wood volume estimation, and others. We introduce an approach leveraging Conditional Generative Adversarial Networks (cGANs) for eucalyptus log segmentation in images, incorporating specialized image processing techniques to handle noise and intersections, coupled with the Connected Components Algorithm for efficient counting. To support this research, we created and made publicly available a comprehensive database of 466 images containing approximately 13,048 eucalyptus logs, which served for both training and validation purposes. Our method demonstrated robust performance, achieving an average Accuracypixel of 96.4% and Accuracylogs of 92.3%, with additional measures such as F1 scores ranging from 0.879 to 0.933 and IoU values between 0.784 and 0.875, further validating its effectiveness. The implementation proves to be efficient with an average processing time of 0.713 s per image on an NVIDIA T4 GPU, making it suitable for real-time applications. The practical implications of this method are significant for operational forestry, enabling more accurate inventory management, reducing human errors in manual counting, and optimizing resource allocation. Furthermore, the segmentation capabilities of the model provide a foundation for advanced applications such as eucalyptus stack volume estimation, contributing to a more comprehensive and refined analysis of forestry operations. The methodology’s success in handling complex scenarios, including intersecting logs and varying environmental conditions, positions it as a valuable tool for practical applications across related industrial sectors.

Improved particle filter algorithm combined with culture algorithm for collision Caenorhabditis elegans tracking

In order to address the issue of tracking errors of collision Caenorhabditis elegans, this research proposes an improved particle filter tracking method integrated with cultural algorithm. The particle filter algorithm is enhanced through the integration of the sine cosine algorithm, thereby facilitating uninterrupted tracking of the target C. elegans. Furthermore, the cultural algorithm is employed to facilitate recognition of the target C. elegans following a collision. In addition, this method integrates the concepts of down-sample and marking to reduce the average processing time of the image. Ultimately, the experiment was conducted on two strains of C. elegans of six ages. The experimental results demonstrate that the proposed method can accurately identify the target worm in the post-collision stage. The proposed method has the potential to be utilized in the field of worm tracking, offering a novel method into the acquisition of collision C. elegans behavior.

State Perception and Health Prediction of Key Equipment in Power Metering Production

Combining intelligent sensing technology with an improved grey wolf optimizer (GWO) algorithm, this article innovatively studied state perception and health prediction methods for key power metering equipment. A state perception network was constructed, utilizing intelligent sensing for real-time collection and monitoring of equipment operation. Gabor transform was applied for signal denoising, followed by empirical mode decomposition to extract operational features. An improved GWO algorithm enhanced search efficiency and global optimization, avoiding local optima. A combination prediction model was established to improve prediction accuracy and reliability. Results showed average data collection and processing times of 0.053 and 0.196 seconds, with 93.84% accuracy, and fault warning and false alarm rates of 93.37% and 6.63%, respectively. This method effectively monitors equipment state and predicts health, aiding proactive maintenance planning and reducing fault costs.

COMPUTED TOMOGRAPHIC ANGIOGRAPHY DATA SET CONTAINING ABDOMINAL AORTIC SEGMENTATION

BACKGROUND: Artificial Intelligence (AI) is widely actively in radiology. The effectiveness of AI algorithms essentially depends on the availability of relevant and representative training and test datasets. Currently, such datasets are highly demanded in open access, particularly CT angiographic (CTA) studies of the abdominal aorta containing not only pathology classification but also vessel segmentation. The disadvantages of existing solutions include small sample volume, narrow specialization, and disparate data preparation methodology. AIM – Design of a CTA dataset containing abdominal aortic segmentation for cases of normal, dilatation, thrombosis, and calcinosis. METHODS: In accordance with the methodology of AI-algorithms testing, the terms of reference for datasets preparation were developed, the necessary sample volume was calculated and approval of the independent ethical committee was obtained. Previously designed original algorithm of semi-automatic segmentation of abdominal aorta on CTA-studies using open access software Slicer 3D was used. Inclusion criteria: CTA studies, or abdominal CT studies with contrast enhancement; presence of arterial phase of the scan; slice thickness not more than 3 mm. Exclusion criteria: presence of any foreign objects in the aortic lumen, aortic dissection. The algorithm was tested on patient data obtained from the Unified Radiologic Information System. Expert evaluation of the compliance of the obtained results with the formed requirements, as well as the assessment of time costs when using the developed segmentation algorithm was carried out. RESULTS: The calculated sample size was 100 CTA studies with a slice thickness ≤ 1.2 mm. Population data: the number of unique patients was 100, the proportion of female patients was 51%; the median age was 62 years ranged from 18 to 84 years. The proportion of pathologic studies was 61%, of which (including combined pathology): 60 cases contained calcification, both by 18 cases with aortic dilatation and thrombosed lumen. Different non-overlapping masks were used for each object. The dataset is presented in the public domain. The average processing time is 0.8 hours per 100 CT-images. CONCLUSION: The public dataset containing 100 CTA studies of segmented abdominal aorta for normal, dilated, thrombotic and calcified cases was created. Results are valuable for AI-algorithms design and anthropomorphic 3D-modeling.

YOLOv8-Based XR Smart Glasses Mobility Assistive System for Aiding Outdoor Walking of Visually Impaired Individuals in South Korea

This study proposes an eXtended Reality (XR) glasses-based walking assistance system to support independent and safe outdoor walking for visually impaired people. The system leverages the YOLOv8n deep learning model to recognize walkable areas, public transport facilities, and obstacles in real time and provide appropriate guidance to the user. The core components of the system are Xreal Light Smart Glasses and an Android-based smartphone, which are operated through a mobile application developed using the Unity game engine. The system divides the user’s field of vision into nine zones, assesses the level of danger in each zone, and guides the user along a safe walking path. The YOLOv8n model was trained to recognize sidewalks, pedestrian crossings, bus stops, subway exits, and various obstacles on a smartphone connected to XR glasses and demonstrated an average processing time of 583 ms and an average memory usage of 80 MB, making it suitable for real-time use. The experiments were conducted on a 3.3 km route around Bokjeong Station in South Korea and confirmed that the system works effectively in a variety of walking environments, but recognized the need to improve performance in low-light environments and further testing with visually impaired people. By proposing an innovative walking assistance system that combines XR technology and artificial intelligence, this study is expected to contribute to improving the independent mobility of visually impaired people. Future research will further validate the effectiveness of the system by integrating it with real-time public transport information and conducting extensive experiments with users with varying degrees of visual impairment.

Developing an automatic treatment record review system for quality assurance of patient treatment delivery in radiation therapy

Background and purposeTreatment record contains most of information related to treatment plan delivery in radiation therapy. Reviewing treatment record is an important quality assurance (QA) task for safety and quality of patient treatments. This task is usually performed by senior medical physicists. However, it is time-consuming, tedious, and error-prone. To assist this process, a treatment record review system (TRRS) is developed to automatically review items related to treatment delivery record.MethodsThe treatment record is firstly extracted from oncology information system (OIS). Based on the daily patient treatment information, the original plan from the treatment planning system is identified. Then the original plan and the delivered plan are correlated. Eight review categories (parameter consistency, treatment completeness, treatment progression, image guidance, override, treatment couch, documentation, and treatment mode) are created. Tailored rules are designed for various review items to automate the review process. As a result, for each daily treatment record, a reviewed flag (pass, failure, warning, and N/A) is assigned by the TRRS. Finally, this system is evaluated by 6 months patient treatment records collected in our institute and compared to the manual process on the same data.ResultsTRRS processed a total of 76,651 treatment fractions from 4230 patients with an average of 574 treatments per day. The percentage of the detected anomalies among the total records was 0.76%. The average processing time was 3.9 s and 282 s per treatment record for the automatic and manual processes, respectively. Comparing with the manual process, the time efficiency of TRRS is improved by a factor of 72. The average numbers of anomalies detected by the automatic and manual processes are 21 and 13 per day, respectively. TRRS detects 61.5% more anomalies than those of the manual process.ConclusionTRRS is not only efficient in processing a large amount of treatment records on a daily basis but also effective in finding more anomalies than those of physics weekly check. The application of the TRRS could significantly reduce the workload of the review physicists and let them focus on more important works related to patient safety.

NNFit: A Self-Supervised Deep Learning Method for Accelerated Quantification of High- Resolution Short Echo Time MR Spectroscopy Datasets.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop and evaluate the performance of NNFit, a self-supervised deep-learning method for quantification of high-resolution short echo-time (TE) echo-planar spectroscopic imaging (EPSI) datasets, with the goal of addressing the computational bottleneck of conventional spectral quantification methods in the clinical workflow. Materials and Methods This retrospective study included 89 short-TE whole-brain EPSI/GRAPPA scans from clinical trials for glioblastoma (Trial 1, May 2014-October 2018) and major-depressive-disorder (Trial 2, 2022- 2023). The training dataset included 685k spectra from 20 participants (60 scans) in Trial 1. The testing-dataset included 115k spectra from 5 participants (13 scans) in Trial 1 and 145k spectra from 7 participants (16 scans) in Trial 2. A comparative analysis was performed between NNFit and a widely used parametric-modeling spectral quantitation method (FITT). Metabolite maps generated by each method were compared using the structural- similarity-index-measure (SSIM) and linear-correlation-coefficient (R2). Radiation treatment volumes for glioblastoma based on the metabolite maps were compared with the Dice-coefficient and a two-tailed t test. Results Average SSIM and R2 scores for Trial 1 test set data were 0.91/0.90 (choline), 0.93/0.93 (creatine), 0.93/0.93 (n-acetylaspartate), 0.80/0.72 (myo-inositol), and 0.59/0.47 (glutamate + glutamine). Average scores for Trial 2 test set data were 0.95/0.95, 0.98/0.97, 0.98/0.98, 0.92/0.92, and 0.79/0.81 respectively. The treatment volumes had average Dice coefficient of 0.92. NNFit's average processing time was 90.1 seconds, whereas FITT took 52.9 minutes on average. Conclusion This study demonstrates that a deep learning approach to spectral quantitation offers comparable performance to conventional quantification methods for EPSI data, but with faster processing at short-TE. ©RSNA, 2025.

An Intelligent Ensemble Deep Learning Techniques with Improved Owl Search Algorithm-Aided Optimal Feature Selection for Predicting the Presence of Heart Diseases

Cardiovascular disease evaluation and prediction are critical medical duties to ensure precise classification, which allows heart specialists to offer treatment to patients. Heart illness is one of the deadliest diseases, particularly a sudden cardiac arrest, which strikes an individual so suddenly that there is no opportunity to manage it, and this condition is extremely hard to detect. Numerous medical machine learning, as well as data mining methods, have been employed to extract significant information about coronary artery disease prognosis. However, the precision of the intended outcomes is not adequate. An effective deep learning-assisted heart disease prediction model is presented to solve the above difficulties. At first, the data are accumulated from a standard dataset. After this process, the features from the data are optimally selected using the Improved Owl Search Algorithm (IOSA). These features are further transferred to the prediction phase. Here, the heart disease is predicted by the Ensemble Deep Attention with Convolution Network (EDACNet), and this model is composed by the combination of “Long Short-Term Memory (LSTM)”, “Gated Recurrent Unit (GRU)”, “Deep Markov Random Field (Deep MRF)” and “1Dimensional Convolutional Neural Network (1DCNN)”. The prediction scores obtained from each structure are multiplied by an optimized weight, and it is subjected to a weighted averaging process. Here, the weight is optimized by the same IOSA. The final results after the weighted averaging process provide the final predicted outcomes over heart diseases. At last, the validation process is carried out to measure the usefulness of a designed prediction model. The result of the developed model achieves better performance in terms of accuracy and precision. It shows the value of accuracy to be 96 and also the precision attained the value of 97. Therefore, the entire validation developed a model that provides effective performance over existing techniques.

A New Deep Learning-Based Method for Automated Identification of Thoracic Lymph Node Stations in Endobronchial Ultrasound (EBUS): A Proof-of-Concept Study.

Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is a cornerstone in minimally invasive thoracic lymph node sampling. In lung cancer staging, precise assessment of lymph node position is crucial for clinical decision-making. This study aimed to demonstrate a new deep learning method to classify thoracic lymph nodes based on their anatomical location using EBUS images. Bronchoscopists labeled lymph node stations in real-time according to the Mountain Dressler nomenclature. EBUS images were then used to train and test a deep neural network (DNN) model, with intraoperative labels as ground truth. In total, 28,134 EBUS images were acquired from 56 patients. The model achieved an overall classification accuracy of 59.5 ± 5.2%. The highest precision, sensitivity, and F1 score were observed in station 4L, 77.6 ± 13.1%, 77.6 ± 15.4%, and 77.6 ± 15.4%, respectively. The lowest precision, sensitivity, and F1 score were observed in station 10L. The average processing and prediction time for a sequence of ten images was 0.65 ± 0.04 s, demonstrating the feasibility of real-time applications. In conclusion, the new DNN-based model could be used to classify lymph node stations from EBUS images. The method performance was promising with a potential for clinical use.

Automated CAD system for early detection and classification of pancreatic cancer using deep learning model

Accurate diagnosis of pancreatic cancer using CT scan images is critical for early detection and treatment, potentially saving numerous lives globally. Manual identification of pancreatic tumors by radiologists is challenging and time-consuming due to the complex nature of CT scan images and variations in tumor shape, size, and location of the pancreatic tumor also make it challenging to detect and classify different types of tumors. Thus, to address this challenge we proposed a four-stage framework of computer-aided diagnosis systems. In the preprocessing stage, the input image resizes into 227 × 227 dimensions then converts the RGB image into a grayscale image, and enhances the image by removing noise without blurring edges by applying anisotropic diffusion filtering. In the segmentation stage, the preprocessed grayscale image a binary image is created based on a threshold, highlighting the edges by Sobel filtering, and watershed segmentation to segment the tumor region and we also implement the U-Net method for segmentation. Then refine the geometric structure of the image using morphological operation and extracting the texture features from the image using a gray-level co-occurrence matrix computed by analyzing the spatial relationship of pixel intensities in the refined image, counting the occurrences of pixel pairs with specific intensity values and spatial relationships. The detection stage analyzes the tumor region’s extracted features characteristics by labeling the connected components and selecting the region with the highest density to locate the tumor area, achieving a good accuracy of 99.64%. In the classification stage, the system classifies the detected tumor into the normal, pancreatic tumor, then into benign, pre-malignant, or malignant using a proposed reduced 11-layer AlexNet model. The classification stage attained an accuracy level of 98.72%, an AUC of 0.9979, and an overall system average processing time of 1.51 seconds, demonstrating the capability of the system to effectively and efficiently identify and classify pancreatic cancers.

Real-time data processing for serial crystallography experiments.

We report the use of streaming data interfaces to perform fully online data processing for serial crystallography experiments, without storing intermediate data on disk. The system produces Bragg reflection intensity measurements suitable for scaling and merging, with a latency of less than 1 s per frame. Our system uses the CrystFEL software in combination with the ASAP::O data framework. In a series of user experiments at PETRA III, frames from a 16 megapixel Dectris EIGER2 X detector were searched for peaks, indexed and integrated at the maximum full-frame readout speed of 133 frames per second. The computational resources required depend on various factors, most significantly the fraction of non-blank frames (`hits'). The average single-thread processing time per frame was 242 ms for blank frames and 455 ms for hits, meaning that a single 96-core computing node was sufficient to keep up with the data, with ample headroom for unexpected throughput reductions. Further significant improvements are expected, for example by binning pixel intensities together to reduce the pixel count. We discuss the implications of real-time data processing on the `data deluge' problem from recent and future photon-science experiments, in particular on calibration requirements, computing access patterns and the need for the preservation of raw data.

Functional Central Limit Theorems for epidemic models with varying infectivity and waning immunity

We study an individual-based stochastic epidemic model in which infected individuals gradually become susceptible again following each infection (generalized SIS model). The epidemic dynamics is described by the average infectivity and susceptibility processes in the population together with the numbers of infected and susceptible/uninfected individuals. In R. Forien et al., Stochastic epidemic models with varying infectivity and susceptibility. arXiv preprint arXiv:2210.04667 (2022), a functional law of large numbers (FLLN) is proved as the population size goes to infinity, and asymptotic endemic behaviors are also studied. In this paper, we prove a functional central limit theorem (FCLT) for the stochastic fluctuations of the epidemic dynamics around the FLLN limit. The FCLT limit for the aggregate infectivity and susceptibility processes is given by a system of stochastic non-linear integral equation driven by a two-dimensional Gaussian process.