Very-high-resolution remote-sensing imagery provides a scalable basis for delineating informal settlements, but sparse annotations, severe imbalance between informal-settlement and background pixels, and cross-city heterogeneity in urban morphology and image–mask correspondence complicate model development. We present SLUM-i, a semi-supervised semantic segmentation framework together with a geographically diverse seven-city Earth observation benchmark spanning three continents. The benchmark combines a newly annotated Lahore dataset and companion Karachi and Mumbai datasets with four publicly released city datasets, totaling 14,458 RGB image–mask tiles. We quantify cross-city heterogeneity using class composition, boundary morphology, grayscale separability, correspondence between mask boundaries and image edges, and divergence between measured feature distributions. For label-efficient mapping, SLUM-i combines representation-guided unlabeled-pool curation, using embeddings from a vision foundation model (DINOv2-Small) to remove the least-similar tiles, and Class-Aware Adaptive Thresholding, which adapts pseudo-label acceptance by class through global mean-confidence and per-class mean-softmax exponential moving averages. Experiments at 10

TCAV (Testing with Concept Activation Vectors) is an interpretability method that assesses the alignment between the internal representations of a trained neural network and human-understandable, high-level concepts. Though effective, TCAV suffers from significant computational overhead, inter-layer disagreement of TCAV scores, and statistical instability. This work takes a step toward addressing these challenges by introducing E-TCAV, a framework for efficient approximation of TCAV scores, which is based on extensive investigation into three key aspects of the TCAV methodology: 1) the effect of latent classifiers on the stability of TCAV scores, 2) the inter-layer agreement of TCAV scores, and 3) the use of the penultimate layer as a fast proxy for earlier layers for TCAV computation. To ensure a solid foundation for E-TCAV, we conduct extensive evaluations across four different architectures and five datasets, encompassing problems from both computer vision and natural language domains. Our results show that the layers in the final block of the neural network strongly agree with the penultimate layer in terms of the TCAV scores, and the commonly observed variance of the TCAV scores can be attributed to the choice of the latent classifier. Leveraging this inter-layer agreement and the degeneracy of directional sensitivities at the penultimate layer, E-TCAV guarantees linearly scaling speed-ups with respect to the network's size and the number of evaluation samples, marking a step towards efficient model debugging and real-time concept-guided training.

OBJECTIVES:Traditional forensic identification relies on DNA, fingerprints, and dental records, which may be unavailable in degraded remains. This study investigates palatal rugae as an alternative intraoral marker and evaluates a fully automated digital matching process with the hypothesis that it can achieve superior accuracy compared with semi-automated methods. METHODS:Palatal scans from 345 participants (11.2-73.2 years) were recorded using an intraoral scanner. After segmentation and Poisson disk downsampling (10,000 points), a four-step matching process was applied: (1) data preparation, (2) rough alignment via Fast Point Feature Histograms (FPFH), (3) fine alignment using the Iterative Closest Point (ICP) algorithm, and (4) distance-based similarity scoring. The algorithm was validated on 224 development scans and 111 unseen scans. Matching accuracy and sampling effects were analyzed. RESULTS:The automated system achieved 100% accuracy in identifying identical scans and rejecting non-matching pairs. Random rotation confirmed robustness to initial alignment. Downsampling analysis showed that 3000 points suffice for perfect accuracy, reducing computational demand without loss of precision. CONCLUSIONS:Automated 3D alignment of palatal rugae enables highly accurate, reproducible, and objective identification. The method outperforms semi-automated approaches and could significantly accelerate forensic workflows in mass casualty events. Further validation with post-mortem and scanner-diverse datasets is required, but results support integrating digital palatal data into standardized forensic identification protocols. CLINICAL SIGNIFICANCE:Automated 3D analysis of palatal rugae enables fast, reproducible and operator-independent identification from digital intraoral scans, supporting forensic workflows when conventional identifiers are unavailable.

Integrating cell segmentation with tracking is critical for achieving a detailed and dynamic understanding of cellular behavior. This integration facilitates the study and quantification of cell morphology, movement, and interactions, offering valuable insights into developmental processes, drug response, and disease mechanisms. Traditional segmentation and tracking methods rely on labor-intensive annotations—such as full segmentation masks or bounding boxes for every cell in each frame—which severely limit throughput and scalability. PointTrack addresses these challenges by introducing a weakly-supervised pipeline that leverages sparse point annotations in the initial frame to automatically generate high-fidelity cell masks and maintain consistent object identities across subsequent frames. PointTrack begins with multiple sparse point annotations per cell in the first image of a time-lapse sequence. An interactive segmentation engine uses these annotations to delineate accurate cell boundaries, eliminating the need for exhaustive mask drawing. A dedicated point-tracking module then propagates the annotated points through the entire video, adapting to changes in cell appearance and motion with minimal user intervention. By reducing annotation effort to a few clicks per cell while preserving mask quality and identity continuity, PointTrack significantly accelerates the preparation of large-scale microscopy datasets. Extensive evaluation on two diverse time-lapse datasets—CTMC and DeepCell—demonstrates robust segmentation and reliable tracking performance under varying imaging modalities, cell densities, and noise conditions. Qualitative examples highlight precise delineation of overlapping cells and seamless trajectory recovery through transient occlusions. With streamlined annotation requirements and automated processing, PointTrack offers an effective and scalable solution for large-scale microscopy-based cell segmentation and tracking.

Deep neural networks are among the most successful algorithms in terms of performance and scalability across different domains. However, since these networks are black boxes, their usability is severely restricted due to a lack of interpretability. Existing interpretability methods do not address the analysis of time-series-based networks specifically enough. This paper shows that an analysis in the frequency domain can not only highlight relevant areas in the input signal better than existing methods but is also more robust to fluctuations in the signal. In this paper, FreqAtt is presented - a framework that enables post-hoc interpretation of time-series analysis. To achieve this, the relevant frequencies are evaluated, and the signal is either filtered or the relevant input data is marked. FreqAtt is evaluated using a wide range of statistical metrics to provide a broad overview of its performance. The results show that using frequency-based attribution, especially in combination with traditional attribution on top of the frequency-optimized signal, provides strong performance across different metrics.

Effective document intelligence models rely on large amounts of annotated training data. However,procuring sufficient and high-quality data poses significant challenges due to the labor-intensive andcostly nature of data acquisition. Additionally, leveraging language models to annotate real documents raises concerns about data privacy. In this context, synthetic document generation has emergedas a promising, privacy-preserving alternative. We propose DocDjinn, a novel framework for controllable synthetic document generation using Vision-Language Models (VLMs) that produces annotated documents from unlabeled seed samples. Our approach generates visually plausible and semantically consistent synthetic documents that follow the distribution of an existing source dataset through clustering-based seed selection with parameterized sampling. By enriching documents with realistic diffusion-based handwriting and contextual visual elements via semantic-visual decoupling, we generate diverse, high-quality annotated synthetic documents. We evaluate DocDjinn as a source of trainingdata across eleven benchmarks spanning key information extraction, question answering, documentclassification, and document layout analysis. Our experiments show that with only 100 real training samples, our framework achieves on average 87% of the performance of the full real-world dataset. To our knowledge, this is the first work demonstrating that VLMs can generate faithful annotated document datasets at scale from unlabeled seeds that can effectively enrich or approximate real, manually annotated data for diverse document understanding tasks. We publicly release our code and 140k+synthetic document samples.

Accurate cell tracking in microscopy is essential for studying biological dynamics like proliferation and migration. Traditional fully supervised methods demand dense pixel-wise masks for every frame, making them impractical for large-scale use. Recent methods like SAT reduce annotation effort by using sparse point-based supervision, but still require multiple positive and negative points per cell, which remains labor-intensive. BoxTrack offers a lightweight and annotation-efficient alternative, requiring only a single bounding box per cell in the first frame. Without relying on any point-level annotations, it performs end-to-end instance segmentation and tracking over entire sequences. This simplification leads to a substantial reduction in annotation cost while improving performance over SAT. On the CTMC dataset, BoxTrack improves Multiple Object Tracking Accuracy (MOTA) by +15.96 https://github.com/nabeelkhalid92/Box-it-Track-it .

Cell segmentation is a cornerstone of biological image analysis, enabling downstream applications such as tracking, counting, and phenotypic profiling. Deep learning has markedly improved segmentation accuracy, leading to broad adoption in research and clinical workflows. However, most current approaches operate on 2D images, neglecting spatial information in the third dimension. This omission is critical, as cellular morphology, migration, and division are inherently three‐dimensional processes. Modern microscopy now offers high‐throughput volumetric imaging, presenting new opportunities for methods that fully exploit 3D data. This review surveys segmentation strategies that incorporate depth information, extending beyond conventional 2D analysis. It distinguishes between pseudo‐3D (2.5D) methods, which balance computational efficiency with limited volumetric context, and fully 3D methods that process volumetric input and output directly. In total, 31 beyond‐2D segmentation approaches are examined and compared, alongside 32 curated volumetric datasets annotated with key metadata such as resolution, size, and ground‐truth availability. Furthermore, a unified reference dataset, compiled from high‐quality open‐source resources, is presented to promote standardized data access, metadata harmonization, and future community‐driven benchmarking efforts in 3D cell segmentation.

Cardiovascular diseases (CVDs) remain the leading cause of mortality worldwide, yet early risk detection is often limited by available diagnostics. Carotid ultrasound, a non-invasive and widely accessible modality, encodes rich structural and hemodynamic information that is largely untapped. Here, we present a machine learning (ML) framework that extracts clinically meaningful representations of vascular damage (VD) from carotid ultrasound videos, using hypertension as a weak proxy label. The model learns robust features that are biologically plausible, interpretable, and strongly associated with established cardiovascular risk factors, comorbidities, and laboratory measures. High VD stratifies individuals for myocardial infarction, cardiac death, and all-cause mortality, matching or outperforming conventional risk models such as SCORE2. Explainable AI analyses reveal that the model relies on vessel morphology and perivascular tissue characteristics, uncovering novel functional and anatomical signatures of vascular damage. This work demonstrates that routine carotid ultrasound contains far more prognostic information than previously recognized. Our approach provides a scalable, non-invasive, and cost-effective tool for population-wide cardiovascular risk assessment, enabling earlier and more personalized prevention strategies without reliance on laboratory tests or complex clinical inputs.

Colorectal polyps are primarily detected through colonoscopy, which plays a central role in early cancer prevention. Precise polyp segmentation supports treatment planning and diagnostic accuracy by providing masks that encode clinically relevant structures. Recent advancements in deep learning have led to several polyp segmentation models. However, performance remains hindered by challenges such as image noise, complex textures, indistinct boundaries, and diverse polyp morphologies. The high cost and time burden of manual annotation underscore the need for automated segmentation systems. To overcome these limitations, BAASNet, a Boundary-Aware Attention-Based Segmentation framework, is introduced for polyp segmentation. A boundary-aware loss function is integrated to improve performance, particularly in delineating polyp edges. The method is evaluated on nine publicly available datasets spanning five imaging modalities, including two center-wise polyp detection benchmarks, demonstrating strong generalization capability. On PolypDB, the model attains a mean Dice similarity coefficient (mDSC) of at least [Formula: see text] across all five modalities. Across all evaluated benchmarks, the proposed model achieves an average absolute improvement of approximately [Formula: see text] in Dice. Gains vary by dataset, ranging from approximately [Formula: see text] to [Formula: see text] relative improvement over the best previous results. These results demonstrate BAASNet's potential for robust, real-time clinical deployment in automated colonoscopy workflows.

Remote sensing imagery provides essential information for various applications. Automated detection of cloud-covered areas in satellite imagery is a crucial first step in many of these applications. Pixel-level identification of clouds in satellite images is a challenging task due to the diverse types of clouds and limited number of bands available in different satellite systems. Recently deep convolutional networks have been often used to achieve state-of-the-art performance in pixel-level object identification for remote sensing applications. In this paper, to address the challenging task of accurate semantic segmentation of clouds in multispectral satellite imagery, we propose an end-to-end attention-based deep convolutional neural network. The architecture of our proposed network comprises an encoder, an attention module, and a decoder. The attention-based model employs a ResNet encoder-decoder backbone, enhanced with residual connections, spatial pyramid pooling, and atrous convolutions. This combination enables the network to focus on more relevant features and improve cloud extraction accuracy. We evaluated our approach on two benchmark datasets from different satellite systems. Experimental results show that the proposed network achieves superior performance compared to existing advanced methods. across several segmentation metrics, including precision, recall, accuracy, and the Jaccard index.

One of the top limiting factors for machine learning applications in industry is the lack of data, as the training of powerful models often requires vast datasets. Acquiring data at large scales is generally expensive and sometimes not even possible, as it may depend on costly measurements with specialized equipment, trained personnel and possibly manual annotations by experts. Thus, augmenting or synthesizing data plays a key role in making machine learning solutions feasible and cost-effective for industrial applications. In this paper, we propose a diffusion model for synthesizing audio data in a real-world multi-label audio classification task from the automotive industry. Augmenting the training dataset with synthetic data, we obtain an improvement of 1.8 p.p. in mAP over training with real data alone. We discuss the difficulties in generating such domain-specific data and examine issues further by comparing our method to an alternative generative approach for environmental sound augmentation.

This study investigates hypertension as a visual marker of individual vascular damage, which can signal an elevated risk of major cardiovascular events. By leveraging machine learning, we aim to identify such damage early and gain insight into a patient’s arterial health. We adapted the VideoMAE deep learning model—originally designed for video classification—by fine-tuning it for ultrasound imaging applications. The model was trained and tested using a dataset comprising over 31,000 carotid sonography videos sourced from the Gutenberg Health Study (15,010 participants), one of the largest prospective population health studies. This adaptation facilitates the classification of individuals as hypertensive or non-hypertensive (74.6

This paper introduces VAEneu, a novel autoregressive method for multistep ahead univariate probabilistic time series forecasting, designed to address the challenges of generating sharp and well-calibrated probabilistic forecasts without assuming a specific parametric form for the predictive distribution. VAEneu leverages the Conditional VAE framework and optimizes the likelihood of the predictive distribution using the Continuous Ranked Probability Score (CRPS), a strictly proper scoring rule, as the loss function. This approach enables the model to learn flexible, sharp, and well-calibrated predictive distributions without the need for a tractable likelihood function. In a comprehensive empirical study, VAEneu is rigorously benchmarked against 12 baseline models across 12 datasets, demonstrating superior performance in both forecasting accuracy and uncertainty quantification. VAEneu provides a valuable tool for quantifying future uncertainties, and our extensive empirical study lays the foundation for future comparative studies for univariate multistep ahead probabilistic forecasting.

Video-based AI systems are increasingly adopted in safety-critical domains such as autonomous driving and healthcare. However, interpreting their decisions remains challenging due to the inherent spatiotemporal complexity of video data and the opacity of deep learning models. Existing explanation techniques often suffer from limited temporal coherence, insufficient robustness, and a lack of actionable causal insights. Current counterfactual explanation methods typically do not incorporate guidance from the target model, reducing semantic fidelity and practical utility. We introduce Latent Diffusion for Video Counterfactual Explanations (LD-ViCE), a novel framework designed to explain the behavior of video-based AI models. Compared to previous approaches, LD-ViCE reduces the computational costs of generating explanations by operating in latent space using a state-of-the-art diffusion model, while producing realistic and interpretable counterfactuals through an additional refinement step. Our experiments demonstrate the effectiveness of LD-ViCE across three diverse video datasets, including EchoNet-Dynamic (cardiac ultrasound), FERV39k (facial expression), and Something-Something V2 (action recognition). LD-ViCE outperforms a recent state-of-the-art method, achieving an increase in R2 score of up to 68

Concept-based explanations have emerged as an effective approach within Explainable Artificial Intelligence, enabling interpretable insights by aligning model decisions with human-understandable concepts. However, existing methods rely on computationally intensive procedures and struggle to efficiently capture complex, semantic concepts. This work introduces the Concept Directions via Latent Clustering (CDLC), which extracts global, class-specific concept directions by clustering latent difference vectors derived from factual and diffusion-generated counterfactual image pairs. CDLC reduces storage requirements by ∼4.6× and accelerates concept discovery by ∼5.3× compared to the baseline method, while requiring no GPU for clustering, thereby enabling efficient extraction of multidimensional semantic concepts across latent dimensions. This approach is validated on a real-world skin lesion dataset, demonstrating that the extracted concept directions align with clinically recognized dermoscopic features and, in some cases, reveal dataset-specific biases or unknown biomarkers. These results highlight that CDLC is interpretable, scalable, and applicable across high-stakes domains and diverse data modalities.