AdaBoost-based satellite detection of summer daytime sea fog and low clouds in ice floe fields of the Arctic

Retraction Note: Staked deep ensemble model for intruder behaviour detection and classification in cloud

Cloud-based telephony platforms face growing fraud risks including voice phishing (vishing), subscription abuse, and organizational impersonation, with detection being especially challenging in low-resource languages such as Arabic. We present an Artificial Intelligence (AI)-based security architecture for fraud detection in Arabic cloud call centers, combining onboarding verification, behavioral monitoring, domain-adapted Automatic Speech Recognition (ASR), semantic transcript search, and Large Language Model (LLM)-based entity verification. The domain-adapted Langa ASR model achieves a Word Error Rate (WER) of 41.0% and Character Error Rate (CER) of 18.2%, outperforming all evaluated commercial baselines. LLM-based entity extraction with multi-call consensus achieves 97.3% company-name accuracy (Generative Pre-trained Transformer 4, GPT-4) and 92.0% in the cost-effective deployed configuration (GPT-3.5 with log-probability filtering). Evaluated on production data from a Middle East and North Africa (MENA)-region provider spanning more than 1000 accounts, the pipeline flagged 47 accounts of which 41 were confirmed fraudulent (directly observed precision 87.2%, 95% confidence interval (CI): 74.3–95.2%; estimated recall 51–82% under conservative base-rate assumptions—not directly measured), providing evidence for the viability of a unified, threat-model-driven architecture for low-resource telephony fraud detection.

Accurate cloud detection remains a significant challenge due to the spectral ambiguity between clouds and bright or heterogeneous surfaces (e.g., snow, desert). While multi-angle and polarization data offer rich information, the discriminative power of joint spectral analysis for resolving these ambiguities has been underexploited. In this work, we demonstrate that physically motivated spectral band ratios and differences can robustly enhance cloud signatures. Motivated by this insight, we propose a novel deep learning framework, the Multi-angle Polarization Feature Pyramid Structure (MP-FPS), that explicitly leverages joint spectral features as discriminative priors. Our architecture employs a dual-branch network to disentangle and adaptively fuse spectral and multi-angle polarization modalities. Within this framework, a hierarchical, multi-scale cross-channel multi-angle fusion module dynamically captures spatial–spectral–angular dependencies, enriching the structural representation of clouds. Furthermore, a channel-space dual-path attention mechanism refines sub-pixel responses, significantly improving detection accuracy in challenging regions such as cloud edges and thin cirrus. Evaluated on the global POLDER-3 dataset, MP-FPS achieves a mean Intersection over Union (mIoU) of 0.8662 across diverse surface types, surpassing the official baseline by 12.4%. This study establishes joint spectral analysis as a critical enabler for high-precision cloud masking, and demonstrates its synergistic value when integrated with multi-angle polarimetric information in a unified deep architecture.

Because of the recent exponential rise in attack frequency and sophistication, the proliferation of smart things has created significant cyber security challenges. Even though the tremendous changes cloud computing has brought to the business world, its centralization makes it challenging to use distributed services like security systems. Valuable data breaches might occur due to the high volume of data that moves between businesses and cloud service suppliers, both accidental and malicious. Unlike outsiders, insiders possess privileged and proper access to information and resources. In this work, a machine learning-based system for insider threat detection and classification is proposed and developed a systematic approach to identify various anomalous occurrences that may point to anomalies and security problems associated with privilege escalation. By combining many models, ensemble learning enhances machine learning outcomes and enables greater prediction performance. Multiple studies have been presented regarding detecting irregularities and vulnerabilities in network systems to find security flaws or threats involving privilege escalation. But these studies lack the proper identification of the attacks. This study proposes and evaluates ensembles of Machine learning (ML) techniques in this context. This paper implements machine learning algorithms for the classification of insider attacks. A customized dataset from multiple files of the CERT dataset is used. Four machine learning algorithms, i.e., Random Forest (RF), Adaptive boosting(AdaBoost), Extreme Gradient Boosting(XGBoost), and Light Gradient Boosting Machine(LightGBM), are applied to that dataset and analyzed results. Overall, LightGBM performed best.

Clouds significantly affects the power output of solar energy systems, and it also decrease the life of modules in photovoltaic system and receivers in the concentrating solar power system, resulting in costs of operation and maintenance. In the present study, the MRR-YOLO model was proposed, which was based on deep learning and instance segmentation technique. The MSDA, the RCS-OSA, and the RFAConv modules were used, and their functions to the cloud segmentation were investigated. Results found that the MSDA module helped maintain the model's lightweight, the RFAConv module had a better feature extraction of the clouds. The instance segmentation method was better than semantic segmentation in clouds detection, especially for clouds of varying shapes. The PB, the RB, and the mAP50B of the MRR-YOLO model were 79.2%, 66%, and 74.7%, respectively. For the segmentation task, the PM, the RM, and the mAP50M were 79.3%, 64.8%, and 73%, respectively. The MRR-YOLO model was validated by the SWIMSEG, the CCSN, the all-sky datasets, and the real cloud images in Zhengzhou city, it had a better detection performance and applicability than other models. A heatmap comparison was also conducted, showing that the model accurately detected all features of the clouds.

Detection of a Molecular Cloud toward the Heartbeating Gamma-Ray Source near the Microquasar SS 433

TLNet: A deep learning framework for tree detection in forest point clouds using multi-layered forest structure

RIFLe-Net: Rotation Invariant Feature Learning Network towards affordance detection in 3D point clouds

KeyGeoFusion: A multi-modal keypoint and geometry-aware framework for small and distant 3D object detection in sparse point clouds

Abstract Atmospheric conditions are typically categorized as either clear‐sky or cloudy‐sky scenarios, yet a distinct Cloud‐Clear Sky Transition Zone (CCTZ) inherently exists. Current studies on CCTZ identification and its shortwave radiative effects remain constrained by limited applicability over land and insufficient consideration of cloud‐type differentiation. To bridge this gap, we integrated Moderate Resolution Imaging Spectroradiometer (MODIS) satellite observations (1 km spatial resolution) with radiative transfer modeling to develop a globally consistent CCTZ detection method applicable to both land and ocean. We subsequently analyzed its effects on direct and diffuse shortwave radiation across seven major cloud types. Key findings include: (a) The CCTZ exhibits a globally coherent but cloud‐type‐dependent spatial scale, with a statistical boundary located approximately 5 km from clouds. The frequency of CCTZ occurrence reaches ∼34% in “no‐cloud” regions over land, significantly higher than over ocean (∼21%). (b) Analysis of shortwave radiation responses, incorporating cloud classification, revealed three distinct mechanisms corresponding to distance from the nearest cloud: Increasing‐then‐decreasing (Stratocumulus‐Stratus, Cirrus), Fluctuating (Altostratus‐Altocumulus, Cumulus, Cumulonimbus), and Stable (Nimbostratus, Cirrus‐Altocumulus‐Altostratus). (c) Compared to pure clear‐sky conditions, the CCTZ enhances global mean diffuse shortwave radiation by ∼16.3%, while reducing direct shortwave radiation by ∼0.8%, with magnitudes modulated by cloud type. This study clarifies the globally coherent yet cloud‐type‐modulated characteristics of the CCTZ and its distinct radiative mechanisms. These findings underscore the need for advanced cloud detection and more refined modeling of transition zones to better constrain their representation in climate systems, thereby helping to reduce uncertainties in global climate predictions.

Accurate cloud detection is an important preprocessing step for subsequent remote sensing data processing. Traditional threshold cloud detection methods have a complex process and require a large number of threshold tests. In recent years, deep learning has been widely applied to cloud detection. However, annotating training datasets for deep learning models typically requires substantial human effort and time investment. Consequently, there are few existing manually annotated cloud detection datasets, and MODIS cloud detection datasets are particularly scarce. To overcome this limitation, we proposed a cloud detection method that combines radiative transfer simulations with deep learning. We first produced a simulated cloud detection dataset using a radiative transfer model and some existing remote sensing products, and then proposed a neural network for training the cloud detection model. Compared with other deep learning models for cloud detection, our method has achieved satisfactory results on the simulated dataset overall. Furthermore, we conducted cloud detection experiments on real satellite imagery. For comparative analysis, we trained other deep learning models on a real satellite image dataset and compared their performance with that of models trained on our simulated dataset. The cloud detection results on real satellite images demonstrate that the models trained on the simulated dataset we proposed achieve performance comparable to those trained on real remote sensing datasets. Specifically, for MODIS data, we compared our results with the official MODIS cloud mask product, MOD35. The results indicate that our method achieves lower false detection rates on mixed surfaces of snow and bare land.

Cloud detection serves as a critical preprocessing step in remote sensing image processing and quantitative applications. However, prevailing deep learning-based models often depend on computationally intensive backbone networks to achieve high accuracy, which hinders their deployment in resource-constrained scenarios such as on-board processing or edge computing. To bridge the trade-off between accuracy and efficiency, this paper introduces a lightweight network for cloud detection. The core innovations of our network are twofold: (1) a dual-path feature enhancer that operates at the front end to extract and fuse multi-scale features through a parallel architecture, significantly enriching feature diversity and representational capacity, thereby alleviating the need for a complex backbone, and (2) a bidirectional gated fusion module, which adaptively integrates multi-scale features from the dual-path feature enhancer with deep semantic features from the backbone decoder through a gated attention mechanism and dynamic convolution, thereby enhancing feature discriminability. Comprehensive experiments on the public HRC_WHU dataset demonstrate that the proposed model achieves a high overall accuracy of 96.31% and a mean intersection-over-union of 92.82%, with only 12.04 GFLOPs of computational cost, outperforming several state-of-the-art methods. These results validate that our approach effectively balances high detection performance with computational efficiency, offering a practical solution for real-time, lightweight cloud detection in high-resolution remote sensing imagery.

In in-vehicle intelligent driving systems, although LiDAR–camera fusion can acquire 3D information, it remains insufficient for detecting small distant objects beyond 50 m due to sparse point clouds and limited feature representation. Existing detection methods mostly focus on close-range targets and do not adequately correspond to the challenges of small pixel occupancy, sparse point cloud, and easy confusion of small targets in the distant view. We propose a Frus-PointPillars-based algorithm that projects 2D detection boxes into 3D frustums to narrow the search scope and suppress background, employs dual pooling in the voxel feature extractor to retain both global and local information, and integrates a multi-scale residual graph-convolution fusion module with a global-aware attention mechanism. On KITTI, our model achieves 68.78%/64.17%/56.43% 3D mAP for Car/Cyclist/Pedestrian and 47.01% long-range (|x|≥40m) mAP (+3.81% versus PointPillars). On nuScenes, 3D mAP rises by 5.28% at 31FPS, demonstrating enhanced accuracy and robustness for distant small-object detection.

Raising the level of abstraction is considered key to addressing the ever-increasing complexity of embedded system design, but it causes additional challenges due to the larger abstraction gap between the initial specification and the final implementation. This article addresses the current lack of systematic design methods by extending existing design-transformation-based approaches and wrapping them into a rule-based transformational design methodology for heterogeneous multi-processor platforms. The methodology cross-fertilizes embedded system design with program transformation techniques while taking into account the interplay of tight constraints and platform heterogeneity inherent in such systems. It advocates step-wise transformations starting from initial requirements to yield a final refined model that is efficient for implementation. To consider the effect of transformations on different properties of the system at each step, the system is specified with a set of requirements, an application model, a platform model, and a set of mapping decisions; referred to as the RAMP view of the system. The RAMP view and its carefully selected underlying unified abstract graph representation lay the foundations for mechanizing and potentially automating design transformations. A pattern matching technique is introduced and a proof-of-concept tool is implemented that automatically detects all possible transformations by matching the patterns defined by the transformation rules to the abstract graph representation of the system model. The underlying graph representation enables complex transformations on different aspects of the design, resulting in an improved design space definition. The design space can be explored by application-platform co-exploration techniques, yielding the most promising sequence of application transformations alongside the best matching platform. The applicability and potential of the proposed methodology are showcased through the design of both an image processing system and a cloud detection system.

Few-shot semantic segmentation for clearance intrusion risk detection in metro tunnel point clouds

Reinforced model selection for resource efficient anomaly detection in edge clouds

Point-HN: Unified spatial-context modeling for fast and accurate 3D point cloud detection

Enhancing cloud detection across multiple satellite sensors using a combined Swin Transformer and UPerNet deep learning model

The unavoidable presence of clouds and their shadows in optical satellite imagery hinders the true spectral response of the Earth’s underlying surface. Accurate cloud and cloud shadow detection is therefore a crucial preprocessing step for optical satellite images and any downstream analysis. Various methods have been developed to address this critical task and can be broadly categorized into physical rule-based methods and learning based methods. In recent years, machine learning based methods, particularly deep learning frameworks, have proven to outperform physical rule-based models. However, these approaches are mostly fully supervised and require a large amount of pixel-level annotations whose acquisition is costly and time consuming. In this work, we propose to address cloud and cloud shadow detection in optical satellite images using self-supervised representation learning, a machine learning paradigm that focuses on extracting relevant representations from unlabeled data, which can then be used as an effective starting point to fine-tune models with few labeled data in a supervised fashion. These approaches have been shown to perform competitively with fully supervised methods without the requirement of large annotation datasets. Specifically, we assessed two self-supervised representation learning methods that use different philosophies about self-supervision: Momentum Contrast (MoCo), based on contrastive learning and DeepCluster, based on clustering. Using two publicly available Sentinel-2 cloud datasets, namely WHUS2–CD+ and CloudSEN12, we show that MoCo and DeepCluster, trained with only 25 % of the annotated data, can perform better than physical rule-based methods such as FMask and Sen2Cor, weakly supervised methods and even several fully supervised methods. These results highlight the strong applicability of self-supervised representation learning methods to the task of cloud and cloud shadow detection with self-supervised pretraining leading to fine-tuned models that outperform industry standards and achieve near state-of-the-art performance with a fraction of the data. • Two self-supervised approaches are leveraged for cloud and cloud shadow detection. • MoCo and DeepCluster outperformed industry standards such as FMask and Sen2Cor with few labels. • Reliable performance was obtained for both methods using 25 % of annotations in the training sets. • MoCo and DeepCluster handle clouds better than cloud shadows.