Results 121 to 130 of about 28,509 (291)

Upsampling DINOv2 Features for Unsupervised Vision Tasks and Weakly Supervised Materials Segmentation

Advanced Intelligent Systems, EarlyView.
Feature from recent image foundation models (DINOv2) are useful for vision tasks (segmentation, object localization) with little or no human input. Once upsampled, they can be used for weakly supervised micrograph segmentation, achieving strong results when compared to classical features (blurs, edge detection) across a range of material systems.
Ronan Docherty, Antonis Vamvakeros, Samuel J. Cooper   +2 more
wiley   +1 more source

Global Optimization strategies for two-mode clustering [PDF]

Two-mode clustering is a relatively new form of clustering that clusters both rows and columns of a data matrix. To do so, a criterion similar to k-means is optimized.
Castilli, W., Groenen, P.J.F., Rosmalen, J.M. van, Trejos, J.   +3 more
core   +1 more source

TacEva: A Performance Evaluation Framework for Vision‐Based Tactile Sensors

Advanced Intelligent Systems, EarlyView.
This work introduces TacEva, a unified framework for evaluating vision‐based tactile sensors. It standardizes intrinsic, performance, and robustness metrics through shared experimental procedures and links them to task‐level requirements in robotic manipulation.
Qingzheng Cong, Steven Oh, Wen Fan, Shan Luo, Kaspar Althoefer, Dandan Zhang   +5 more
wiley   +1 more source

KDLM: Lightweight Brain Tumor Segmentation via Knowledge Distillation

Advanced Intelligent Systems, EarlyView.
A lightweight student network is designed, which is based on multiscale and multilevel feature fusion and combined with the residual channel attention mechanism to achieve efficient feature extraction and fusion with very few parameters. A dual‐teacher collaborative knowledge distillation framework is proposed.
Baotian Li, Haoyu Tang, Jun Zhu, Fang Fang Gou, Jia Wu   +4 more
wiley   +1 more source

Deep Learning Methods for Assessing Time‐Variant Nonlinear Signatures in Clutter Echoes

Advanced Intelligent Systems, EarlyView.
Motion classification from biosonar echoes in clutter presents a fundamental challenge: extracting structured information from stochastic interference. Deep learning successfully discriminates object speed and direction from bat‐inspired signals, achieving 97% accuracy with frequency‐modulated calls but only 48% with constant‐frequency tones. This work
Ibrahim Eshera, Gaurav Duggal, Rolf Müller   +2 more
wiley   +1 more source

Artificial Intelligence in Autonomous Mobile Robot Navigation: From Classical Approaches to Intelligent Adaptation

Advanced Intelligent Systems, EarlyView.
Artificial intelligence (AI) is reshaping autonomous mobile robot navigation beyond classical pipelines. This review analyzes how AI techniques are integrated into core navigation tasks, including path planning and control, localization and mapping, perception, and context‐aware decision‐making. Learning‐based, probabilistic, and soft‐computing methods
Giovanna Guaragnella, Angelo Cardellicchio, Cosimo Patruno, Massimiliano Nitti, Nicola Pedrocchi, Vito Renò   +5 more
wiley   +1 more source

Entropy‐Guided Convolutional Neural Network Classification of Sensor Signals for Real‐Time Surface Quality Monitoring in Direct Laser Interference Patterning

Advanced Intelligent Systems, EarlyView.
Neural‐network pipeline for real‐time DLIP surface‐quality monitoring: spectral entropy of WLI topographies is used to generate interpretable K‐means labels, which are transferred to time‐resolved photodiode traces. A compact dual‐input 1D‐CNN (signal + laser parameters) learns discriminative spatiotemporal features and predicts “OK/NOK” surface ...
Marcelo Daniel Sallese, Ignacio Tabares, Wei Wang, Marcos Soldera, Andrés Fabián Lasagni   +4 more
wiley   +1 more source

Digital Surface‐Enhanced Raman Scattering With Event Counting and Spectrum Learning for Label‐Free Protein Quantification

Advanced Intelligent Systems, EarlyView.
A statistical and machine learning‐assisted surface‐enhanced Raman scattering (SERS) framework is developed for label‐free quantification of low‐abundance analytes, including proteins. Combining digital SERS event counting with binomial regression and an artificial neural network (ANN) trained on full spectra, the approach achieves picomolar detection ...
Eni Kume, James Rice
wiley   +1 more source

Ion‐Lock Storage With Multi‐Logic Circuitry Gated by Polar–Dipolar Interactions in Poly(Ionic Liquids)

Angewandte Chemie, EarlyView.
Precise building block design of poly(ionic liquids) creates an opportunity to achieve ionic conductivity, energy storage, and multi‐value logic in one material while eliminating elaborate fabrication processes. ABSTRACT In these studies, we developed a new generation of polymeric materials capable of electrical energy storage and forming higher‐than ...
Sourav Biswas, Jiahui Liu, Jack Austin, Marek W. Urban   +3 more
wiley   +2 more sources

RT‐DETR‐DA for Complex Scenes: Distracted Driving Detection With Feature Interaction and Dynamic Perception

Advanced Intelligent Systems, EarlyView.
This work proposes RT‐DETR‐DA, an enhanced real‐time detection framework for identifying distracted driving in complex, real‐world environments. The model introduces a dynamic sparse gating multiscale attention module and an attention‐guided dual‐path fusion module to strengthen multiscale perception and cross‐layer feature interaction.
Yi Liu, QiaoXing Li, XiBin Wang, Ying He, Sen Zhang   +4 more
wiley   +1 more source