Hong Kong University of Science and Technology
Hong Kong University of Science and Technology Institutional RepositoryNot a member yet
162821 research outputs found
Sort by
TextRSR: Enhanced Arbitrary-Shaped Scene Text Representation Via Robust Subspace Recovery
No full textIn recent years, scene text detection research has increasingly focused on arbitrary-shaped texts, where text representation is a fundamental problem. However, most existing methods still struggle to separate adjacent or overlapping texts due to ambiguous spatial positions of points or segmentation masks. Besides, the time efficiency of the entire pipeline is often neglected, resulting in sub-optimal inference speed. To tackle these problems, we first propose a novel text representation method based on robust subspace recovery, which robustly represents complex text shapes by combining orthogonal basis vectors learned from labeled text contours. These basis vectors capture basis contour patterns with distinct information, enabling clearer boundaries even in densely populated text scenarios. Moreover, we propose a dynamic sparse assignment scheme for positive samples that adaptively adjusts their weights during training, which not only accelerates inference speed by eliminating redundant predictions but also enhances feature learning by providing sufficient supervision signals. Building on these innovations, we present TextRSR, an accurate and efficient scene text detection network. Extensive experiments on challenging benchmarks demonstrate the superior accuracy and efficiency of TextRSR compared to state-of-the-art methods. Particularly, TextRSR achieves an F-measure of 88.5% at 37.8 frames per second (FPS) for CTW1500 dataset and an F-measure of 89.1% at 23.1 FPS for Total-Text dataset.</p
Overcoming Barriers and Solutions for Catalysing Private Capital in Climate Adaptation: A Stakeholder-Informed Agenda for Hong Kong's Intermediary Role in Southeast Asia
No full textThis study investigates stakeholder perspectives on mobilising private-sector finance for climate adaptation in Southeast Asia, emphasising Hong Kong's role as a financial intermediary. Through semi-structured interviews with diverse stakeholders, including practitioners, policymakers, insurers, and project developers, we employed a grounded theory approach to identify key themes. The findings reveal significant opportunities for private capital across five adaptation domains: flood resilience, energy resilience, early warning systems, digital infrastructure resilience, and insurance coverage. However, stakeholders also identified barriers, such as fragmented project pipelines, the public-good nature of adaptation initiatives, and challenges in quantifying adaptation risks. To overcome these obstacles, participants advocated for a Hong Kong-based Adaptation Aggregation Facility, developing blended-finance instruments, and introducing an adaptation taxonomy with disclosure protocols to enhance local capacity. Policy implications highlight the need for seed capital, credit enhancements, tax incentives, and regulatory sandboxes to attract private investment. This study provides actionable insights for scaling private adaptation finance and underscores priorities for further interdisciplinary and practice-oriented research
Hygroscopicity and Cloud Condensation Nucleation Activity of Bromine Salts
No full textHygroscopicity of bromine salt influences the formation of cloud condensation activity (CCN) and CCN number concentrations, thereby affecting cloud formation. However, the current study on the hygroscopicity of bromine-containing aerosols has not been fully understood. In this study, we focus on the hygroscopicity and CCN of bromine salt such as sodium bromide (NaBr) and ammonium bromide (NH4Br) using a hygroscopic tandem differential mobility analyzer (H-TDMA) and a cloud condensation nuclei counter (CCNC) at 24 ± 1 ℃. The hygroscopic behavior of bromine salts differs from humic acid’s hygroscopic growth, with deliquescence relative humidity (DRH) and different CCN for NaBr and NH4Br aerosol particles. The hygroscopic parameter κ for NaBr was determined to be 0.89, larger than the κ value of 0.77 measured for NH4Br. NaBr deliquesced at 45%, compared with a DRH of 75% for NH4Br aerosols. However, 100 nm humic acid aerosols show no obvious DRH. The measurements are accompanied by RH-dependent thermodynamic equilibrium calculations using the Aerosol Inorganic–Organic Mixtures Functional groups Activity Coefficients (AIOMFAC) model. There are several effects of humic acid on the hygroscopicity and CCN activity of mixtures containing NaBr and NH4Br in relation to the different mass fractions of organic compounds: (1) a shift of DRH of NaBr to higher RH (1.5 to 3.7%) with increasing mass fraction from 25 to 50 wt% of humid acid in the mixture. No DRH was observed in the bromine-containing aerosols with an HA mass fraction of 75 wt%. (2) the usage of Zdanovskii–Stokes–Robinson (ZSR) relation leads to agreement with measured diameter growth factors of aerosol particles containing humic acid and bromine salt. (3) the κ differences between sub- and supersaturated conditions ranged from − 0.28 to 0.14 for 1:1 NaBr/HA, 1:3 NaBr/HA, and all NH4Br/HA mixture aerosols, falling within experimental uncertainty. Therefore, it provides insights into the hygroscopicity and cloud condensation nuclei of bromine salts and their interactions with humic acid under sub- and supersaturated conditions, which are crucial for comprehending the atmospheric life processes and impacts of bromine salts.</p
In situ transplantation and multiple omics reveal holobiont adaptation in deep-sea mussel Gigantidas haimaensis
No full textDeep-sea mussels Gigantidas haimaensis rely on methane-oxidizing bacteria (MOB) endosymbionts for nutrition in methane seeps, yet the molecular mechanisms enabling holobiont resilience to environmental fluctuations remain unclear. Here, we integrate a chromosome-scale genome of G. haimaensis with an in situ transplantation experiment and multi-omics analyses to investigate adaptive responses to methane limitation. Transplanting mussels to a low-methane environment for 6 days reduced MOB abundance by 30.6%. Meta-transcriptomics showed that MOB prioritized methane oxidation via upregulated pmoA/pmoB genes but downregulated amino acid biosynthesis and non-essential pathways, indicating metabolic resource reallocation. Concurrently, host transcriptomics revealed a shift from symbiont-dependent strategies (“farming” and “milking”) to filter-feeding and extracellular matrix remodeling, indicating changes in trophic level. This dynamic interplay demonstrates how the holobiont balances symbiont maintenance with alternative energy acquisition under stress and highlights the vulnerability of chemosynthetic symbioses to methane fluctuations induced by environmental changes.</p
Elastodynamic gauge transformation for controlling vibration mode shapes in Willis networks
No full textDesigning mechanical metamaterial networks to precisely control vibration mode shapes at targeted frequencies has shown great potential for applications in fluid and particle manipulation, nondestructive structural health monitoring and evaluation, as well as the protection of sensitive and fragile structures. However, the inverse design of metamaterial networks to simultaneously satisfy vibration mode shape and frequency requirements remains elusive. Here, we suggest a rapid analytical approach, called elastodynamic gauge transformation, to exploit design guidelines on microscopic geometry and local material properties to construct metamaterial networks to achieve sought-after vibration mode shapes at target frequencies. We operate a displacement gauge in the transformation over the Lagrangian of metamaterial networks and build the relationships between the desired displacement field and microscopic design. This approach is exemplified through discrete string, beam, and linear spring networks. We find that transformed string networks keep the same network geometry as those before the transformation, but each of the strings should be grounded and display Willis coupling. Further, beam networks can also conserve their virtual microscopic geometry, but additional grounded torsional springs and their associated rotational Willis coupling must be included. Nevertheless, linear spring networks no longer maintain their virtual network geometry after the transformation, where both the directions of the linear springs and the shapes of the masses need to be properly adjusted. We also interpret the transformed discrete string, beam, and linear spring networks at the continuum limit by performing elastodynamic gauge transformation over homogenized membrane, plate, and Cauchy elasticity models. Numerical simulations are performed to validate the elastodynamic gauge transformation for string, beam, and linear spring networks. Elastodynamic gauge transformation offers a new approach for the rapid design of metamaterial structures to achieve desired vibration mode shapes at target frequencies.</p
Ten-Step Synthesis of Potent Anticancer Steroid Withanolide D
No full textWithanolide D is a highly oxygenated steroid with a broad spectrum of biological activities (potent antitumor properties) and thus has stimulated considerable interest from synthetic chemists. Herein, we describe a concise 10-step synthesis of withanolide D, which features a Diels-Alder reaction with singlet oxygen, followed by a Kornblum-DeLaMare rearrangement to streamline functionalization of the critical A ring. A vinylogous aldol reaction was employed to forge the E ring with high regioselectivity and stereoselectivity. It should be noted that withanolide D could serve as the precursor for syntheses of six additional withanolide-type natural products, including withaferin A. This work establishes an efficient and modular route to withanolide D and related analogues, providing a robust platform for the synthesis of diverse withanolide derivatives as potential anticancer drug candidates.</p
Accurate concrete spalling segmentation from bounding box supervision using Segment Anything
No full textAccurate pixel-level segmentation of concrete spalling has been severely hampered by the prohibitive cost of manual annotation. This paper investigates how accurate pixel-level defect segmentation can be achieved using only low-cost weakly supervised bounding box annotations. A three-stage framework is proposed to generate and refine pseudo-masks from bounding boxes using the Segment Anything Model (SAM), dynamic self-correction, and inference-time fusion. The proposed method outperformed existing techniques by over 10% in F1 score on a large-scale spalling dataset. These findings establish the economic viability of deploying scalable automated inspection systems by drastically reducing data annotation costs, providing a practical and scalable pathway for spalling assessment.</p
TexGS-VolVis: Expressive Scene Editing for Volume Visualization via Textured Gaussian Splatting
No full textAdvancements in volume visualization (VolVis) focus on extracting insights from 3D volumetric data by generating visually compelling renderings that reveal complex internal structures. Existing VolVis approaches have explored non-photorealistic rendering techniques to enhance the clarity, expressiveness, and informativeness of visual communication. While effective, these methods often rely on complex predefined rules and are limited to transferring a single style, restricting their flexibility. To overcome these limitations, we advocate the representation of VolVis scenes using differentiable Gaussian primitives combined with pretrained large models to enable arbitrary style transfer and real-time rendering. However, conventional 3D Gaussian primitives tightly couple geometry and appearance, leading to suboptimal stylization results. To address this, we introduce TexGS-VolVis, a textured Gaussian splatting framework for VolVis. TexGS-VolVis employs 2D Gaussian primitives, extending each Gaussian with additional texture and shading attributes, resulting in higher-quality, geometry-consistent stylization and enhanced lighting control during inference. Despite these improvements, achieving flexible and controllable scene editing remains challenging. To further enhance stylization, we develop image- and text-driven non-photorealistic scene editing tailored for TexGS-VolVis and 2D-lift-3D segmentation to enable partial editing with fine-grained control. We evaluate TexGS-VolVis both qualitatively and quantitatively across various volume rendering scenes, demonstrating its superiority over existing methods in terms of efficiency, visual quality, and editing flexibility.</p
Quantifying the Impact of the Inherent Coupling of Particle-Scale Variability on Granular Soil Behavior: A Micromechanical Study
No full textThis study investigates how the inherent discreteness and coupled variability of particle-scale properties, specifically interparticle friction (μs) and shear modulus (Gp), affect macroscale soil behavior. To achieve a comprehensive probabilistic quantification, copula theory was employed to model the joint dependence between these particle properties, capturing the inherent property uncertainty in granular materials. The results reveal that coupled particle property uncertainties significantly influence macroscale parameters, including the stress ratio, void ratio, and small-strain stiffness, which are strongly associated with desired packing densities and soil states. Notably, the uncertainties of mechanical behaviors converge at the critical state, regardless of packing density, consistent with conventional soil mechanics. In contrast, the impact of coupled μs and Gp on particle-scale stress transmission was negligible, with particle size distribution emerging as the dominant factor. Furthermore, a decoupled multiprobability density evolution method (M-PDEM) was introduced to investigate nonlinear dependencies among macroscale parameters. This approach unveiled observable probabilistic interconnections among stress ratio, void ratio, and coordination number. Such interdependencies, often overlooked in conventional deterministic models, highlight the importance of uncertainty quantification for accurate soil behavior prediction. By integrating copula theory with the decoupled M-PDEM framework, the study offers a robust method for tracing the propagation of microscale uncertainties to macroscale responses, bridging the gap between particle-level variability and engineering-scale behavior. These insights can inform probabilistic geotechnical design methodologies, improve the reliability of numerical simulations, and enhance our understanding of soil behavior under realistic loading and material variability conditions. This study establishes a framework to capture the inherent discreteness and coupling of particle-scale properties, revealing micromechanical mechanisms that govern macroscale soil behavior and strengthening the physical basis for geotechnical analysis and design.</p
Thermal Shielding and Vapor Transport Enhancement in MOF-Enabled Membranes for Membrane Distillation
No full textDesalination via membrane distillation (MD) powered by low-grade or waste heat is an emerging approach to energy-efficient water purification. However, conventional membranes suffer from significant conductive heat loss, which limits their thermal performance. Developing membranes with low thermal conductivity is crucial for enhancing thermal efficiency. This study introduces a metal–organic framework-enabled membrane (MEM) by embedding zeolite imidazole framework 8 (ZIF-8) onto Poly(vinylidene fluoride-co-hexafluoropropylene) (PH) nanofiber. The MEM features a hierarchical porous structure, with an ultralow thermal conductivity (0.03 W m−1 K−1), thereby minimizing heat dissipation. It outperforms conventional membranes, demonstrating a vapor flux of 44.5 LMH and a thermal efficiency of 71.3%, improving MD performance. Multi-scale simulations reveal that the dual improvements in thermal shielding and vapor flow facilitation enable the MEM to effectively harness low-grade heat sources inaccessible to traditional membranes, positioning it as a promising solution for a sustainable water-energy-environment nexus.</p