No modality left behind: Adapting to missing modalities via knowledge distillation for brain tumor segmentation.

CNN and monocular vision-based natural feature matching for monitoring displacement of segmental retaining wall under real-world lighting conditions and non-stationary camera

Reference-based DWI super-resolution network using dual split factor attention and progressive feature matching

M2OTCA: Multiple-magnification optimal transport-based cross-attention learning for whole slide image classification.

Accurate geolocalization of ground targets from unmanned aerial vehicles (UAVs) is critically limited by pose estimation errors and the scarcity of active ranging sensors. To address these challenges, we propose a pipeline that integrates reference image cropping, robust cross-view matching, and geographic projection to estimate real-world coordinates using 2.5D reference maps. For evaluation, we introduce SkyPin, the first large-scale benchmark of its kind, designed to comprehensively test UAV-based localization methods. It comprises UAV imagery from eight diverse environments, featuring both visible and thermal infrared modalities under a wide range of conditions, including variations in weather, time of day, flight altitude, and camera perspective. All ground targets are annotated with centimeter-accuracy Real-Time Kinematic (RTK) coordinates. We establish a comprehensive benchmark by evaluating a series of feature matching methods combined with different projection strategies, allowing systematic comparison of algorithm performance. Representative results show that RoMa combined with PnP-based raytracing achieves the best overall performance, reaching a median 2D error of 0.87 m and Recall@5m values of 0.94 and 0.98 on RGB and thermal infrared UAV-map settings, respectively. Further analysis reveals that performance degrades in challenging mountainous scenes and under large viewing-angle variations, highlighting terrain relief and UAV perspective changes as remaining critical challenges for robust target geo-localization. The full dataset and implementation code will be made publicly available to facilitate future research in UAV-based geolocalization.

Absolute rotation estimation is an important topic in 3D computer vision. Existing works in literature generally employ a multi-stage (at least two-stage) estimation strategy where multiple independent operations (feature matching, two-view rotation estimation, and rotation averaging) are implemented sequentially. However, such a multi-stage strategy inevitably leads to the accumulation of the errors caused by each involved operation, and degrades its final estimation on global rotations accordingly. To address this problem, we propose an End-to end method for estimating Absolution Rotations from multi view images based on deep neural Networks, called EAR-Net. The proposed EAR-Net consists of an epipolar confidence graph construction module and a confidence-aware rotation averaging module. The epipolar confidence graph construction module is explored to simultaneously predict pairwise relative rotations among the input images and their corresponding confidences, resulting in a weighted graph (called epipolar confidence graph). Based on this graph, the confidence-aware rotation averaging module, which is differentiable, is explored to predict the absolute rotations. Thanks to the introduced confidences of the relative rotations, the proposed EAR-Net could effectively handle outlier cases. Experimental results on three public datasets demonstrate that EAR-Net outperforms the state-of-the-art methods by a large margin in terms of both accuracy and inference speed.

Distance-based selection has been the conventional criterion for weather file selections, although this topographically- and meteorologically-agnostic decision may lead to unreliable Climate-Based Daylight Modelling (CBDM) simulation results. To mitigate this risk, this study explores alternative criteria that may be better predictors of weather file proxies for locations within the tropical climate. Principal Component Analysis (PCA) and k-means clustering were conducted on screened tropical climate files to determine the primary attributes and meteorological clusters. Four Principal Components (PCs) explained 89.1% of the variance in weather files. Sensitivity analysis of Malaysian files determined that the four PCs provided a 78% probability of simulating Spatial Daylight Autonomy (sDA) and Useful Daylight Illuminance (UDI) within a 10% tolerance of the Petaling Jaya (PJ) benchmark. Screening for files within weighted PCA distance (PCAd) outperformed pure geometric proximity selections when evaluated at both Malaysian and Class A levels. Overall, the Climatic Similarity Index (CSI) protocol achieved a 86.67% predictive selection accuracy rate across a proportionate random sampling of all Class A TMYx files. This study proves that a combination of atmospheric and geometric feature matching has the potential to overcome the limitations of local weather file availability for a target site.

Theta band activity during event-file retrieval is influenced by stimulus salience in the preceding action episode.

Protocol for a SPIRIT extension for reporting Pragmatic and Explanatory trial protocols designed using the PRECIS tool. (SPIRIT-PRECIS).

Information and Communication Technology (ICT) offers transformative opportunities for educating students with intellectual disabilities. Research shows that ICT tools (e.g., assistive devices, educational software, virtual reality) help learners acquire functional and adaptive skills, increase independence in daily activities, and promote social inclusion. Digital game–based learning and multimedia applications engage students and support understanding, communication, and motivation. Teachers report that ICT benefits diverse learning dimensions (e.g., comprehension, behavior, metacognition, memory). By providing adaptive curricula and learning opportunities, ICT can improve students’ quality of life and vocational prospects. Evidence suggests that systematic ICT training helps retain skills and independence and can empower learners toward lifelong learning opportunities. Furthermore, ICT can broaden educational horizons and reduce inequities by supporting self-directed study and skill development. However, effective use depends on matching technology features to individual needs and applying universal design principles to ensure accessibility. Educators must assess how ICT features align with each student’s needs to maximize benefits. Despite its promise, research on ICT usability for learners with intellectual disabilities remains very limited. In India, inclusive policies and initiatives (e.g., PM eVidya) illustrate efforts to leverage ICT in supporting learners with disabilities. However, challenges such as inadequate infrastructure and insufficient teacher training persist and must be addressed. Moreover, a coordinated effort among educators, policymakers, and technologists is needed to fully harness ICT’s potential in special education. Overall, the literature indicates substantial potential for ICT to empower learners with intellectual disabilities, underscoring the importance of inclusive design, teacher training, and policy support.

Underwater images suffer from wavelength dependent attenuation and multiple scattering, which often lead to severe color casts, veiling effects, and reduced contrast. These degradations weaken the stability of key vision modules such as feature detection, feature matching, edge extraction, and object recognition, and ultimately compromise applications including visual navigation, structural inspection, and environmental monitoring for underwater robots. To improve global color consistency while preserving local texture details, we propose a Physics-Guided Texture-Aware Fusion for Real-World Underwater Image Enhancement (GPRF-HPNet). In the preprocessing stage, a YCbCr domain attenuation map is exploited to guide color correction, followed by entropy driven dual histogram global contrast enhancement. The resulting intermediate images are further combined through gradient weighted wavelet fusion, which retains structural information and fine scale details. High frequency Gabor texture maps at four orientations, namely :0^circ:,45^circ:,90^circ: and 135°, are then constructed as an explicit detail prior. These maps feed a texture branch that runs in paralle with a base branch focusing on structure and color. A parallel residual fusion unit performs joint feature extraction on the two branches, learns adaptive weights, and produces fused feature representations, after which a lightweight decoder reconstructs the enhanced image. Extensive experiments on the Color-Check7, Test-C60 and UCCS datasets demonstrate that the proposed method achieves consistent gains on six metrics, including UIQM and UCIQE and delivers more reliable performance in downstream tasks such as geometric rotation estimation and edge detection, while demonstrating strong generalization across diverse underwater scenes.

Local feature matching plays a critical role in robotic SLAM and visual localization. However, in weakly textured indoor industrial environments, lightweight appearance-based methods often struggle to learn discriminative and stable local features. To address this challenge, this paper proposes GAEFeat, short for Geometry-Aware Efficient Feature, a lightweight vision-geometric feature learning network. To address the scarcity of specialized training data, we integrated robotic arm pose priors with depth information to automatically generate cross-view supervision signals and surface-normal labels. Based on this strategy, we constructed two complementary datasets, including a simulated dataset and a real-world dataset, to support feature learning and evaluation in weakly textured indoor industrial environments. For feature extraction, we design a dual enhancement mechanism consisting of a geometric auxiliary branch and a geometry-aware enhancement (GAE) module. The former guides the network to perceive local surface structures through surface normal supervision, while the latter utilizes a gating mechanism to achieve deep fusion between geometric priors and 2D texture descriptors. Experimental results demonstrate that GAEFeat achieves strong robustness and high inference efficiency in relative pose estimation, homography estimation, and visual localization tasks, with particularly notable advantages in near-field, weakly textured industrial scenes. The framework achieves an inference latency of only 3.9 ms on the NVIDIA Jetson AGX Orin edge platform, demonstrating its real-time capability and practical potential for deployment in edge computing environments.

Video object detection in sports scenarios faces severe challenges posed by high speed motion, heavy occlusion, and complex deformation. Existing methods employ passive feature aggregation that applies a uniform fusion strategy across all regions, neglecting the type specific differences of local degradation and failing to match targeted complementary information for different defects. Meanwhile, visual degradation causes object boundaries to exhibit probabilistic distribution characteristics, yet conventional deterministic regression ignores such geometric ambiguity and forces fitting a single coordinate, making localization reliability difficult to guarantee. To address these issues, we propose ARUDet (Active Retrieval Uncertainty-aware Detector), comprising an Active Temporal Retrieval Module (ATRM) and an Uncertainty Rectified Regression Head (URH). Specifically, ATRM first identifies the degradation type of each region in the current frame through a Defect Responsive Assessor and encodes it into an explicit query vector. Subsequently, under dual constraints of semantic consistency and quality complementarity, it actively retrieves the best matching historical feature patches from a temporal memory bank and selectively replaces only low quality regions, achieving on demand restoration. URH collaboratively suppresses geometric uncertainty through a Probabilistic Boundary Projector (PBP) and a Lower Bound Optimizer (LBO): the former explicitly models the geometric distribution to improve localization quality, while the latter constrains the worst case error to tighten the prediction lower bound. Experiments on multiple sports datasets demonstrate that ARUDet achieves significant performance improvements, further validating the effectiveness of the proposed method.

Traditional fingerprint identification primarily relies on the number of matching minutiae between the questioned and reference prints, where identity is determined by whether the match count exceeds a fixed threshold. However, this "minimum matching pair threshold method" lacks statistical validation. This study establishes a large-scale fingerprint data analysis framework based on artificial intelligence and machine learning to quantify the discriminative value of fingerprint minutiae. A YOLOv12 hybrid model is designed for high-precision minutiae detection. Using 619,338 fingerprints from four pattern classes, the occurrence frequencies of six types of minutiae are statistically analyzed, providing the foundation for quantitative modeling. A minutiae matching method is further proposed to perform large-scale matching according to specified minutia types and quantities. The approach integrates three modules: positional matching, local ridge-flow similarity comparison, and image similarity evaluation. 772 million pairs of non-mated fingerprints are analyzed to empirically refine traditional identification methods and quantify feature-level discriminative value. Results indicate that fingerprint matching stability depends not only on minutiae count but also on fingerprint pattern and minutia types. Building upon Shannon's information theory, a quantitative model for evaluating minutiae discriminative value is established based on feature frequency and matching performance. After residual correction, the model achieves a coefficient of determination (R2) of 0.958, demonstrating high explanatory power and robustness.

Precise reconstruction of plant phenotypes is crucial for smart agriculture. Conventional methods struggle with low efficiency and strong dependency on high-quality data, especially for low-texture and structurally complex crops like wheat. We propose a novel 3D reconstruction framework-Plant3R-that fuses deep feature learning with 3D Gaussian Splatting (3DGS). It innovatively uses the Matching and Stereo 3D Reconstruction (MASt3R) model for sparse point cloud reconstruction and camera pose estimation via its 3D feature matching capabilities, which substantially improve image matching rates and the quality of sparse point clouds. Subsequently, 3DGS is employed for rendering and optimization, enabling end-to-end, high-fidelity, and high-robust 3D reconstruction of wheat plants. Validated on potted wheat at multiple growth stages using handheld images, our experimental results demonstrate that Plant3R performs well in feature extraction and matching, and the reconstructed point cloud provides a good geometric prior for the subsequent rendering stage. In most scenes, its key rendering metrics-Peak Signal-to-Noise Ratio (PSNR) > 34, Structural Similarity Index Measure (SSIM) of 0.94, and Learned Perceptual Image Patch Similarity (LPIPS) < 0.26-surpassed Neural Radiance Fields (NeRF) and the original 3DGS. Moreover, extracted phenotypic traits such as plant height, leaf length, and width showed high correlation with manual measurements (R2 > 0.94), confirming its utility for accurate and quantitative phenotype analysis. Overall, Plant3R not only improves the rendering quality and geometric precision of 3D modeling, but also provides a reliable tool for accurate phenotypic parameter extraction and high-throughput crop phenotyping in precision agriculture.

Deformable image registration plays an essential role in various medical image tasks. Existing deep learning-based deformable registration frameworks primarily utilize convolutional neural networks (CNNs) or Transformers to learn features to predict the deformations. However, the lack of semantic information in the learned features limits the registration performance. Furthermore, the similarity metric of the loss function is often evaluated only in the pixel space, which ignores the matching of high-level anatomical features and can lead to deformation folding. To address these issues, in this work, we proposed LDM-Morph, an unsupervised deformable registration algorithm for medical image registration. LDM-Morph integrated features extracted from the latent diffusion model (LDM) to enrich the semantic information. Additionally, a latent and global feature-based cross-attention module (LGCA) was designed to enhance the interaction of semantic information from LDM and global information from multi-head self-attention operations. Finally, a hierarchical metric was proposed to evaluate the similarity of image pairs in both the original pixel space and latent-feature space, enhancing topology preservation while improving registration accuracy. Extensive experiments on four public 2D cardiac image datasets, two 3D image datasets, show that the proposed LDM-Morph framework outperformed existing state-of-the-art CNNs-and Transformers-based registration methods regarding accuracy with comparable topology preservation and computational efficiency. Our code is publicly available at: https://github.com/wujiong-hub/LDM-Morph.

3D SA-LoIFM: A three-dimensional framework for segmentation-assisted learning of informative feature matching in liver CT-3D US rigid registration

This study presents a verifiable framework for high-fidelity colorless image processing by integrating depth-guided chrominance transfer with multi-level feature extraction. To address critical limitations of traditional methods-color inconsistency, edge bleeding, and insufficient feature matching accuracy-we establish a dual-channel processing pipeline for RGB-D images. This study presents a high-fidelity colorless image processing framework through depth-guided chrominance transfer and multi-level feature extraction. The method introduces three key innovations: a weighted non-local Laplacian algorithm for cross-modal consistency, a multi-level feature hierarchy bridging pixel statistics to semantic understanding, and a brightness-guided edge preservation mechanism. Rigorous testing on benchmark datasets (Middlebury Stereo, MIT-Adobe FiveK, BSDS500, Flickr Million) under standardized preprocessing confirms its efficacy. The algorithm reduces color reconstruction error to 0.020 ± 0.003-a 60% improvement over the Welsh method-with a Turbo-pixel 800 edge fitting accuracy of 0.93 and over-segmentation of only 8%. The three-level feature system boosts matching recall to 91.4%±2.1% from 62.3%, while cutting semantic alignment error from 12.4 to 3.1 ± 0.7 pixels. On MIT-Adobe FiveK, it achieves a PSNR of 38.5 ± 0.4 dB, SSIM of 0.942 ± 0.008, and 9.4 FPS, outperforming diffusion-based colorization by 3.7 dB PSNR with 37% faster speed. All results are statistically significant (p < 0.05) and fully reproducible. This work provides a transparent, efficient, and robust solution for visual graphic design, establishing a new benchmark for colorless image processing.

Liquid chromatography–mass spectrometry (LC–MS) and nuclear magnetic resonance (NMR) spectroscopy are complementary analytical techniques widely used in proteomics, metabolomics, and structural biology. Both generate high-dimensional, noisy spectra where overlapping peaks complicate interpretation. LC–MS relies on retention time (RT) separation before mass analysis, while multidimensional NMR spreads information across chemical-shift axes to reduce congestion. However, comparative or replicate experiments often introduce RT shifts in LC–MS or frequency shifts in NMR, hindering accurate matching of corresponding features. In some experiments, such as variable-temperature NMR, the shifts are intentionally triggered, and frequency tracking provides important information. In any case, a robust, scalable alignment across runs is critical for reliable compound identification, quantification, and structural analysis. We propose a truncated Wasserstein distance-based algorithm for aligning LC–MS and NMR spectra. By constraining maximum transport distance and formulating alignment as a minimum-cost flow problem solved via the Network Simplex algorithm, our method accelerates computation, suppresses spurious matches, and improves robustness to noise. On benchmark LC–MS datasets, it achieved 0.97 precision, 0.96 recall, and a 0.6-s runtime, outperforming OpenMS and DeepRTAlign tools. For NMR data, the algorithm proved effective in 2D, 4D, and even 7D analyses. The algorithm is implemented in wnetalign with supporting modules wnet and pylmcf, available on PyPI and GitHub under permissive licenses: https://github.com/michalsta/pylmcf, https://github.com/michalsta/wnet, https://github.com/michalsta/wnetalign.

With the advancement of visual sensing systems, computer vision is playing an increasingly important role in autonomous driving and robot navigation. Relative pose estimation in multi-camera systems is essential for accurate vehicle localization and environment perception, demanding high real-time performance and robustness. Existing methods, however, often involve high computational costs and rely heavily on abundant feature matches, limiting their applicability in time-sensitive driving scenarios. To address these limitations, this paper introduces a unified framework for efficient relative pose estimation, built upon a novel, to our knowledge, translation parameterization and first-order rotation approximation. Within this framework, we propose three efficient minimal solvers specifically designed for autonomous vehicles. The first solver integrates the vertical direction prior from inertial measurement units (IMUs), the second utilizes the rotation axis direction prior during steering maneuvers, and the third is designed for planar motion-a realistic assumption for ground vehicles operating on structured roads. By reducing both the minimal number of point correspondences and the algebraic complexity, our methods enable faster hypothesis generation within RANSAC-based pipelines, improving suitability for real-time systems. Extensive experiments on synthetic datasets and the KITTI autonomous driving benchmark demonstrate that the proposed solvers achieve a favorable balance between speed and accuracy compared to existing state-of-the-art algorithms.