DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
A comprehensive study on optimization techniques for AMR robots recognition models
No full textIn the process of computing on digital computers and related devices, problems arise in the representation, operation, and storage of numerical values. Quantization and pruning are common methods to address these issues. Using quantization, real-valued numbers can be converted to discrete numbers, thus reducing
computational costs. In recent years, neural network models(such as image classification, object detection, semantic segmentation, natural language processing, etc.)have developed rapidly in many fields. However, in specific domains such as computer vision and natural language processing, neural network models often have high computational requirements, making the quantization process essential. Quantization technology can compress parameters by converting floating point parameters in the model into low-precision integer parameters. This reduces the storage space required for the model and the time it takes to load the model. Autonomous Mobile Robots (AMR) represent a practical application of recognition models, where model compression becomes essential during deployment due to constraints imposed by hardware limitations. This report investigates the core techniques utilized in optimizing neural network models and transformer models for the deployment of AMR systems, with a specific emphasis on quantization and pruning methodologies, and widely adopted optimization frameworks.Master's degre
Ethnic diversity in consistent text-to-image generation
No full textPresent generative AI models successfully create a wide variety of high-quality images but face challenges in personalized and consistent visual story generation. This study evaluates the training-free ConsiStory approach for generating image sets with and without ethnic diversity modifiers in the text prompts. This is achieved by comparing datasets curated for this study: a Benchmark Dataset without diversity keywords and a Diverse Dataset with ethnic groups specified in the prompts. The datasets have 100 corresponding sets of prompts and the images generated with them. DreamSim was used to evaluate the subject consistency in the set of images, while CLIPScore measured the performance in terms of image alignment to the text prompt. Quantitative experiments revealed if there is a statistically significant difference between the performance of corresponding sets in Diverse and Benchmark Datasets. Furthermore, qualitative evaluation is conducted to investigate the composition of the generated image sets in relation to the prompts, which can act as heuristics for downstream tasks. The findings contribute to advancing practical and socially aware text-to-image generative AI applications.Bachelor's degre
Curvature-guided depletion stabilizes Kagome superlattices of nanocrystals
No full textShape-anisotropic nanocrystals and patchy particles have been explored to construct complex superstructures, but most studies have focused on convex shapes. We report that nonconvex, dumbbell-shaped nanocrystals (nanodumbbells) exhibit globally interlocking self-assembly behaviors governed by curvature-guided depletion interactions. By tailoring the local curvature of nanodumbbells, we can precisely and flexibly adjust particle bonding directionality, a level of control rarely achievable with conventional convex building blocks. These nanodumbbells can undergo long-range ordered assembly into various intricate two-dimensional superlattices, including the chiral Kagome lattice. Theoretical calculations reveal that the Kagome lattice is a thermodynamically stable phase, with depletion interactions playing a crucial role in stabilizing these non-close-packed structures. The emergence of Kagome lattices and other unusual structures highlights the vast potential of nonconvex nanocrystals for creating sophisticated architectures.Ministry of Education (MOE)National Research Foundation (NRF)This work was supported by National Key Research and Development Program of China (2022YFA1503501), NSFC (52333013, 22025501, 22088101, 52202088, 52373204, 52394272, 224B1019), Science and Technology Commission of Shanghai Municipality (24TS1410200, 23JC1400100, 2024ZDSYS02, 24JD1400700), Shanghai International Science and Technology Partnership Project (23520750400), China Postdoctoral Science Foundation (2024M760493), and the Fundamental Research Funds for the Central Universities (20720220010). T.L. was supported by the Shanghai Rising-Star Program (24QA2700800). R.N. was supported by the Academic Research Fund from the Singapore Ministry of Education (RG151/ 23 and MOE2019-T2-2-010) and the National Research Foundation, Singapore, under its 29th Competitive Research Program (CRP) Call (NRF-CRP29-2022-0002)
Homogenization and approximation of solutions to multiscale problems
No full textThis project explores how to model steady-state diffusion in media that vary on
very small spatial scales. These types of problems often arise in materials composed of repeating structures, and resolving all the microscopic details can be
difficult. To address this, we use homogenization theory to approximate the behavior of such media by an effective equation that captures the macroscopic response. Starting from a periodic framework, we derive the homogenized equation
using asymptotic expansions and construct the homogenized matrix that governs
large-scale behavior. We then study how to estimate this matrix using both direct
matrix methods and variational principles, leading to classical Voigt–Reiss inequalities. To improve on these estimates, we introduce the Hashin–Shtrikman bounds,
which take into account additional structure like isotropy. The theory is applied
to examples involving composite layers and skew-symmetric flows, showing how
the bounds perform in practice. Overall, the report highlights how homogenization provides a useful tool for connecting fine-scale structure to effective material
properties.Bachelor's degre
Dispossession through decolonisation: tracing the institutional obfuscation of Sabah's history through merger with Malaysia
No full textPerceived as an “equal partner” during the 1963 formation of Malaysia, Sabah’s history has since been subsumed by Malaysian national historical narratives sixty years on. While Sabah’s integration into Malaysia has been characterised as a natural consequence to its decolonisation, such narratives obfuscate the brief period after World War II when Sabah was free to shape its own idealised visions of its postcolonial form. As local elites such as Donald Stephens and Mustapha Harun sought to represent their communities, they were cast to be instrumental in shaping the territory’s post-colonial vision while Sabah was under British rule. However, as political mechanisms such as the Cobbold Commission later justified Sabah’s integration into Malaysia, the same leaders would abandon the interests of its local population, leading them to be utterly dispossessed of their ability to articulate their own regional history. By drawing on the shared interests between the British, the Malayan state, and the local Sabahan elites, this paper explores Sabah’s integration process as a political act that sidelines and secures the political stability of Malaya and the departing British colonial polity, at the expense of its local inhabitants. This paper will reveal Sabah’s integration as an act of political subsumption, that continues into its post-colonial era.Bachelor's degre
Urinary bioorthogonal reporters for the monitoring of the efficacy of chemotherapy for lung cancer and of associated kidney injury
No full textThe utility of urinary tests for the monitoring of the treatment efficacy and adverse events of anticancer therapies is constrained by the low concentration of relevant urinary biomarkers. Here we report, using mice with lung cancer and treated with chemotherapy, of a urinary fluorescence test for the concurrent monitoring of the levels of a tumour biomarker (cathepsin B) and of a biomarker of chemotherapy-induced kidney injury (N-acetyl-β-d-glucosaminidase). The test involves two intratracheally administered urinary reporters leveraging caged bioorthogonal click handles for the biomarker-dependent activation of ‘clickability’ and renal clearance, and the bioorthogonal click reaction of each renally cleared reporter with paired fluorescence indicators in the collected urine. In mouse models of chemotherapy-treated orthotopic lung cancer and of cisplatin-induced kidney injury, lower urinary fluorescence signals (which can be measured by a smartphone camera) for tumour and kidney injury levels positively correlated with animal weight gain and survival time. Biomarker-activated bioorthogonal click reactivity and renal clearance combined with bioorthogonally triggered fluorescence in vitro may enable specific, sensitive and rapid urinary assays for the monitoring of other physiopathological processes.Ministry of Education (MOE)National Research Foundation (NRF)Submitted/Accepted versionK.P. thanks Singapore National Research Foundation (NRF) (NRF-NRFI07-2021-0005), and Singapore Ministry of Education, Academic Research Fund Tier 2 (MOE-T2EP30220-0010; MOE-T2EP30221-0004) for the financial support
Optical resonators constitute a universal spin simulator
No full textNP-hard computational problems can be efficiently recast as finding the ground state of an effective spin model. However, to date no convenient setup exists that can universally simulate all of them, even for a fixed problem size. Here we present such a setup, the universal homogeneous spin simulator (UHSS) using a series of optical (or polaritonic) resonators arranged in a chain using the geometry based on the one introduced in [Phys. Rev. Appl. 21, 024057 (2024)]. We demonstrate by example how the simulator solves Hamiltonian cycle and traveling salesman problems, and show that it generalizes to any NP-hard problem of arbitrary size. This approach works because it allows arbitrary long-range interactions in the spin model to be mapped on an optical system with only nearest-neighbor physical interactions.Ministry of Education (MOE)Published versionWe were supported by the Singaporean Ministry of Education, via the Tier 2 Grant No. MOE-T2EP50121-0020
Electrostatic repellent dispersion method for green and cost-effective aqueous radiative cooling paint
No full textRadiative cooling paint (RCP) serves as an effective energy-saving method for cooling of built infrastructure. Given that the synthesis of most RCPs contains volatile organic compounds (VOC) for binder dispersion, water-based systems have been preferred as cost-effective alternatives that can provide low/no emission of VOC for minimized environmental impact. However, development of cost-effective and robust water-based radiative cooling paint is still challenging. Moreover, there exists particle dispersion issues leading to structural non-uniformity when excess amount of water is used in paints with high particle volume concentration (PVC). Herein, we address these challenges by proposing a universal electrostatic-assisted dispersion method to tackle the dispersion challenge. With ultrahigh PVC (∼70 %), the derived radiative cooling paint exhibits dense and uniform surface while maintaining robust mechanical, chemical, and thermal stabilities, as well as universal adhesivity, significantly contrasting to the cracking and poor adhesivity observed in conventional water dispersed paint. Assisted by superfine dispersion and high PVC, our paint achieves excellent optical properties (Rsolar∼ 97 %, ELWIR> 95 %) for enhanced radiative cooling, and is effective under various climates. This work provides insights into more environmental-friendly and durable designs for radiative cooling paint.Ministry of Education (MOE)Nanyang Technological UniversitySubmitted/Accepted versionThis work was supported by the NTU Gap Fund (NGF-2023-14-014), Tier 1 Project (RG153/23) and MOE-T2EP50123–0016 received by Hong Li
Catalytic reforming of biomass pyrolysis gas over Ni catalysts: alumina, spent fluid catalytic cracking catalyst and char as supports
No full textThe potential replaceability of a Ni catalyst supported on commercial α-Al2O3 (Ni/Al2O3) by Ni on biomass-derived char (Ni/Char), and Ni on spent fluid catalytic cracking catalyst (Ni/FCC) for steam reforming of biomass pyrolysis gas was investigated (14 h at 850°C, steam/carbon ratio = 5). The catalysts reformed 60–80 % of C2-C5 hydrocarbons, producing 2.7–4.1 mg min−1 of H2. The reforming activity of Ni/Al2O3 and Ni/FCC was higher compared to Ni/Char, indicating the beneficial role of metal oxide supports. The use of Al2O3 and FCC resulted in a lower thermo-oxidative stability of coke formed on Ni/Al2O3 and Ni/FCC compared to Ni/Char. Furthermore, the deposited Ni showed higher stability towards oxidation by steam into NiO in case of Al2O3 and FCC compared to char. According to reforming activity, H2 production rate, coking, and Ni oxidation of the catalysts, FCC has better prospects as an alternative support in a reforming catalyst than char.Agency for Science, Technology and Research (A*STAR)National Research Foundation (NRF)Public Utilities Board (PUB)This research is supported by A*STAR under its RIE2025 Industry Alignment Fund- Industry Collaboration Projects (IAF-ICP) Programme (Award I2101E0006). This research is also supported by the National Research Foundation, Singapore, and PUB, Singapore’s National Water Agency under its RIE2025 Urban Solutions and Sustainability (USS) (Water) Centre of Excellence (CoE) Programme which provides funding to the Nanyang Environment & Water Research Institute (NEWRI) of the Nanyang Technological University, Singapore (NTU). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of National Research Foundation, Singapore and PUB, Singapore’s National Water Agency
Real-time detection and auto-focusing of beam profiles from silicon photonics gratings using the YOLO model
No full textWhen observing chip-to-free-space light beams from silicon photonics (SiPh) to free space, manual adjustment of camera lens is often required to obtain a focused image of the light beams. In this Letter, we demonstrated an auto-focusing system based on the you-only-look-once (YOLO) model. The trained YOLO model exhibits high classification accuracy of 99.7% and high confidence level of >0.95 when detecting light beams from SiPh gratings. A video demonstration of real-time light beam detection, real-time computation of beam widths, and auto-focusing of light beams is also included.Agency for Science, Technology and Research (A*STAR)Ministry of Education (MOE)National Research Foundation (NRF)Submitted/Accepted versionAgency for Science, Technology and Research (SPF (C222 517002), NRF2021-QEP2-03-P07, NSTIC (M24W1NS007)); National Research Foundation Singapore (NCAIP (NRF-MSG-2023-0002)); Ministry of Education - Singapore (AcRF Tier 1 (RG135/23, RT3/23)); Ministry of Education - Singapore (AcRF Tier 2 (T2EP50121-0002 (MOE-000180-01)))