Search Results (1,204)

In this work, the utility of multimodal vision–language models (VLMs) for visual product understanding in e-commerce is investigated, focusing on two complementary models: ColQwen2 (vidore/colqwen2-v1.0) and ColPali (vidore/colpali-v1.2-hf). These models are integrated into two architectures and evaluated across various product interpretation tasks, including image-grounded question answering, brand recognition and visual retrieval based on natural language prompts. ColQwen2, built on the Qwen2-VL backbone with LoRA-based adapter hot-swapping, demonstrates strong performance, allowing end-to-end image querying and text response synthesis. It excels at identifying attributes such as brand, color or usage based solely on product images and responds fluently to user questions. In contrast, ColPali, which utilizes the PaliGemma backbone, is optimized for explainability. It delivers detailed visual-token alignment maps that reveal how specific regions of an image contribute to retrieval decisions, offering transparency ideal for diagnostics or educational applications. Through comparative experiments using footwear imagery, it is demonstrated that ColQwen2 is highly effective in generating accurate responses to product-related questions, while ColPali provides fine-grained visual explanations that reinforce trust and model accountability. Full article

This work focuses on the possibilities of generating synthetic data for machine learning in hazard prediction in area-based speed monitoring systems. The purpose of the research conducted was to develop a methodology for generating realistic synthetic data to support the design of a continuous vehicle speed monitoring system to minimize the risk of traffic accidents caused by speeding. The SUMO traffic simulator was used to model driver behavior in the analyzed area and within a given road network. Data from OpenStreetMap and field measurements from over a dozen speed detectors were integrated. Preliminary tests were carried out to record vehicle speeds. Based on these data, several simulation scenarios were run and compared to real-world observations using average speed, the percentage of speed limit violations, root mean square error (RMSE), and percentage compliance. A new metric, the Combined Speed Accuracy Score (CSAS), has been introduced to assess the consistency of simulation results with real-world data. For this study, a basic hazard prediction model was developed using LoRaWAN sensor network data and environmental contextual variables, including time, weather, location, and accident history. The research results in a method for evaluating and selecting the simulation scenario that best represents reality and drivers’ propensities to exceed speed limits. The results and findings demonstrate that it is possible to produce synthetic data with a level of agreement exceeding 90% with real data. Thus, it was shown that it is possible to generate synthetic data for machine learning in hazard prediction for area-based speed control systems using traffic simulators. Full article

This study develops a domain-adaptive multimodal RAG (Retrieval-Augmented Generation) system to improve the accuracy and efficiency of technical question answering based on large-scale structured manuals. Using Hyundai Staria maintenance documents as a case study, we extracted text and images from PDF manuals and constructed QA, RAG, and Multi-Turn datasets to reflect realistic troubleshooting scenarios. To overcome limitations of baseline RAG models, we proposed an enhanced architecture that incorporates sentence-level similarity annotations and parameter-efficient fine-tuning via LoRA (Low-Rank Adaptation) using the bLLossom-8B language model and BAAI-bge-m3 embedding model. Experimental results show that the proposed system achieved improvements of 3.0%p in BERTScore, 3.0%p in cosine similarity, and 18.0%p in ROUGE-L compared to existing RAG systems, with notable gains in image-guided response accuracy. A qualitative evaluation by 20 domain experts yielded an average satisfaction score of 4.4 out of 5. This study presents a practical and extensible AI framework for multimodal document understanding, with broad applicability across automotive, industrial, and defense-related technical documentation. Full article

Tunnel construction poses significant safety challenges due to confined spaces, limited visibility, and the dynamic movement of labourers and machinery. This study addresses a critical gap in real-time, bidirectional proximity monitoring by developing and validating a prototype early-warning system that integrates real-time location systems (RTLS) with long-range (LoRa) wireless communication and ultra-wideband (UWB) positioning. The system comprises Arduino nano microcontrollers, organic light-emitting diode (OLED) displays, and piezo buzzers to detect and signal proximity breaches between workers and equipment. Using an action research approach, three pilot case studies were conducted in a simulated tunnel environment to test the system’s effectiveness in both static and dynamic risk scenarios. The results showed that the system accurately tracked proximity and generated timely alerts when safety thresholds were crossed, although minor delays of 5–8 s and slight positional inaccuracies were noted. These findings confirm the system’s capacity to enhance situational awareness and reduce reliance on manual safety protocols. The study contributes to the tunnel safety literature by demonstrating the feasibility of low-cost, real-time monitoring solutions that simultaneously track labour and machinery. The proposed RTLS framework offers practical value for safety managers and informs future research into automated safety systems in complex construction environments. Full article

To address issues commonly observed during the inference phase of large language models—such as inconsistent labels, formatting errors, or semantic deviations—a series of targeted strategies has been proposed. First, a relation label refinement strategy based on semantic similarity and syntactic structure has been designed to calibrate the model’s outputs, thereby improving the accuracy and consistency of label prediction. Second, to meet the contextual modeling needs of different types of instance bags, a multi-level contextual augmentation strategy has been constructed. For multi-sentence instance bags, a graph-based retrieval enhancement mechanism is introduced, which integrates intra-bag entity co-occurrence networks with document-level sentence association graphs to strengthen the model’s understanding of cross-sentence semantic relations. For single-sentence instance bags, a semantic expansion strategy based on term frequency-inverse document frequency is employed to retrieve similar sentences. This enriches the training context under the premise of semantic consistency, alleviating the problem of insufficient contextual information. Notably, the proposed multi-granularity framework captures semantic symmetry between entities and relations across different levels of context, which is crucial for accurate and balanced relation understanding. The proposed methodology offers practical advancements for semantic analysis applications, particularly in knowledge graph development. Full article

Air quality monitoring systems based on Internet of Things (IoT) technology are critical for addressing environmental and public health challenges, but their energy efficiency poses a significant challenge to their autonomous and scalable deployment. This study investigates strategies to enhance the energy efficiency of IoT-based air quality monitoring systems. A comprehensive analysis of sensor types, data transmission protocols, and system architectures was conducted, focusing on their energy consumption. An energy-efficient system was designed using the Smart Air sensor, Zigbee gateway, and Mini UPS, with its performance evaluated through daily energy consumption, backup operation time, and annual energy use. An integrated efficiency index (IEI) was introduced to compare sensor models based on functionality, energy efficiency, and cost. The proposed system achieves a daily energy consumption of 72 W·h, supports up to 10 h of autonomous operation during outages, and consumes 26.28 kW·h annually. The IEI analysis identified the Ajax LifeQuality as the most energy-efficient sensor, while Smart Air offers a cost-effective alternative with broader functionality. The proposed architecture and IEI provide a scalable and sustainable framework for IoT air quality monitoring, with potential applications in smart cities and residential settings. Future research should explore renewable energy integration and predictive energy management. Full article

The advancement of Internet of Things (IoT) technology has enabled battery-powered devices to be deployed across a wide range of applications; however, it also introduces challenges such as high energy consumption and security vulnerabilities. To address these issues, adiabatic logic circuits offer a promising solution for achieving energy efficiency and enhancing the security of IoT devices. Adiabatic logic circuits are well suited for energy harvesting systems, especially in applications such as sensor nodes, RFID tags, and other IoT implementations. In these systems, the harvested bipolar sinusoidal RF power is directly used as the power supply for the adiabatic logic circuit. However, adiabatic circuits require a peak detector to provide bulk biasing for pMOS transistors. To meet this requirement, a diode-connected MOS transistor-based voltage doubler circuit is used to convert the sinusoidal input into a usable DC signal. In this paper, we propose a novel adiabatic logic design that maintains low power consumption while optimizing energy and current fluctuations across various input transitions. By ensuring uniform and complementary current flow in each transition within the logic circuit’s functional blocks, the design reduces energy variation and enhances resistance against power analysis attacks. Evaluation under different clock frequencies and load capacitances demonstrates that the proposed adiabatic logic circuit exhibits lower fluctuation and improved security, particularly at load capacitances of 50 fF and 100 fF. The results show that the proposed circuit achieves lower power dissipation compared to conventional designs. As an application example, we implemented an ultrasonic transmitter circuit within a LoRaWAN network at the end-node sensor level, which serves as both a communication protocol and system architecture for long-range communication systems. Full article

Traffic measurement systems are an essential part of intelligent transportation systems (ITS). These are specialized transport infrastructures where traffic data is collected and analyzed in order to optimize the use of road systems, improve transport safety, and implement future transport plans. The rapid development of transportation systems, urbanization, and industrialization have led to a global problem of air pollution. This has raised the topical issue of measuring and monitoring environmental parameters at traffic monitoring stations in ITS. In this paper, we present a wireless environmental monitoring system, which is a subsystem of a traffic monitoring station. Along with measuring traffic parameters, the station also collects useful meteorological information. A novel hybrid, dual-band IoT system based on LoRa and LoRaWAN for environmental parameters monitoring is presented. The hardware realization of a developed hybrid LoRaWAN end device, together with the sensors used for the measurement of air parameters, is described. Initial results from real test monitoring of environmental parameters on the road in urban environments are presented as a proof of concept. The presented wireless environmental monitoring system can also be used for indoor or outdoor air pollution monitoring, serving as a useful complement to intelligent transport systems. Full article

Ensuring the safety and quality of perishable agrifood products throughout the supply chain is essential. Key parameters, such as temperature and humidity, must be consistently monitored to prevent spoilage, maintain nutritional value, and minimise health risks. Fluctuations in transportation conditions can compromise product integrity, leading to deterioration and an increased risk of foodborne illness. Monitoring agrifood supply chains is essential, from packaging to last-mile delivery. Distribution methods that rely on non-automated monitoring systems, such as manual temperature measurements, are error-prone due to the failure of manual treatments and increase the likelihood of product deterioration. Emerging sensor technologies and the rapid development of Information and Communication Technologies offer new possibilities for real-time tracking, enabling stakeholders to maintain optimal conditions and monitor aesthetic, physicochemical, and nutritional quality. This paper proposes a cost-effective temperature and humidity traceability system that utilises wireless sensor networks (WSN) and Internet of Things (IoΤ) services to monitor perishable products within the agrifood supply chain ecosystem. It also provides an overview of recent innovations in sensor technologies, along with food quality indicators relevant to real-time monitoring of food quality. The proposed research examines the available sensor technologies and methodologies that enable continuous monitoring of agrifood supply chains. Moreover, the paper presents a pilot full-scale project from both functional and technological perspectives. Full article

The challenge of detecting nuanced, context-dependent offensive language highlights the need for Large Language Model (LLM) adaptation strategies that can effectively address inherent data and task asymmetries. Standard Parameter-Efficient Finetuning (PEFT) methods like Low-Rank Adaptation (LoRA), while efficient, often employ a more uniform, or symmetric, update mechanism that can be suboptimal for capturing such linguistic subtleties. In this paper, we propose Dynamic Sparse LoRA (DS-LoRA), a novel technique that leverages asymmetric adaptation to enhance LLM finetuning for nuanced offensive language detection. DS-LoRA achieves this by (1) incorporating input-dependent gating mechanisms, enabling the asymmetric modulation of LoRA module contributions based on instance-specific characteristics, and (2) promoting asymmetric sparsity within LoRA update matrices via L1 regularization. This dual asymmetric strategy empowers the model to selectively engage and refine only the most pertinent parameters for a given input, fostering a more parsimonious and contextually aware adaptation. Extensive experiments on benchmark datasets demonstrate DS-LoRA’s significant overperformance over standard LoRA and other strong baselines, particularly in identifying subtle and contextually ambiguous offensive content, underscoring the benefits of its asymmetric adaptive capabilities. Full article

Water pollution from pesticides is a major concern for regulatory agencies worldwide due to expensive detecting mechanisms, delays in the processing of results, and the complexity of the chemical analysis. However, the deployment of monitoring systems utilising the internet of things (IoT) and machine-to-machine communication technologies (M2M) holds promise in overcoming this major global challenge. In this current research, an IoT-based wireless sensor network (WSN) is successfully deployed in rural Kenya at the Kiu watershed, providing in situ pesticide detections and a real-time data visualisation of shallow wells. Kiu is an off-grid community located in an area of intensive agriculture, where residents face a high exposure to pesticides due to farming activities and a reliance on shallow wells for domestic water. The evaluation of path loss models utilising channel characteristics obtained from this study indicate a marked departure from the continuous signal decay with distance. Transmitted packets from deployed sensor nodes indicate minimal mutations of payloads, underscoring systems reliability and data transmission integrity. Additionally, the proposed design significantly reduces the time taken to deliver pesticide measurement results to relevant stakeholders. For the entire monitoring period, pesticide residues were not detected in the selected wells, an outcome validated with lab procedures. These results are attributed to prevailing dry weather conditions which limited the leaching of pesticides to lower layers reaching the water table. Full article

The gender gap in drug use is narrowing in regions where access to criminalized substances, such as opioids, is increasing. While research shows that substance use is gendered, less is known about the cultural norms and values shaping women’s drug use, as most studies focus on men. Cross-national comparisons of cultural models of addiction are needed to better understand how addiction is perceived and to inform culturally responsive treatment approaches for women. This study examined cultural models of addiction among reproductive-aged women receiving treatment for substance misuse in London, Toronto, and Delhi. Participants completed a semi-structured questionnaire with open-ended and free-list prompts. Findings revealed shared cultural models attributing drug use to psychological factors, such as self-medicating to manage negative emotions or enhance positive ones, as well as relational, developmental, and biological influences. In conclusion, the study highlights the importance of incorporating cultural models into research and treatment. By using an inductive approach to explore meanings surrounding drug use among people in recovery, researchers can better understand how interventions are received and interpreted through existing internal frameworks. Full article

Long Range (LoRa) is one of the promising Low-Power Wide-Area Network technologies to achieve a strong anti-noise ability due to the modulation of the chirp spread spectrum in low-power and long-distance communications. However, LoRa suffers the problem of packet collisions. Hence, we propose QR−LoRa, a novel PHY-layer scheme that can transmit data in both amplitude and frequency dimensions simultaneously. For the amplitude modulation, we modulate the constant envelope of a LoRa chirp with a cyclic right-shifted ramp signal, where the cyclic right-shifted position carries the data of the amplitude modulation. We adopt the standard LoRa for frequency modulation. We prototype QR−LoRa on the software-defined radio platform USRP N210 and evaluate its performance via simulations and field experiments. The results show the bit rate gain of QR−LoRa is up to 2× compared with the standard LoRa device. Full article

Accurate path loss prediction is essential for optimizing Long-Range Wide-Area Network (LoRaWAN) performance. Previous studies have employed various Machine Learning (ML) models for path loss prediction. However, environmental factors such as temperature, humidity, barometric pressure, and particulate matter have been largely neglected. This study bridges this gap by evaluating the performance of five boosting ML models—AdaBoost, XGBoost, LightGBM, GentleBoost, and LogitBoost—under dynamic environmental conditions. The models were compared with theoretical models (Log-Distance and Okumura-Hata) and existing studies that employed the same dataset based on metrics such as RMSE, MAE, and R². Furthermore, a detailed performance vs. complexity analysis was conducted using metrics such as training time, inference latency, model size, and energy consumption. Notably, barometric pressure emerged as the most influential environmental factor affecting path loss across all models. Bayesian Optimization was applied to fine-tune hyperparameters to improve model accuracy. Results showed that LightGBM outperformed other models with the lowest RMSE of 0.5166 and the highest R² of 0.7151. LightGBM also offered the best trade-off between accuracy and computational efficiency. The findings show that boosting algorithms, particularly LightGBM, are highly effective for path loss prediction in LoRaWANs. Full article

The stylistic emulation of historical building facades presents significant challenges for artificial intelligence (AI), particularly for complex and data-scarce styles like Krakow’s Eclecticism. This study aims to develop a methodological framework for a “typological transcoding” of style that moves beyond mere visual mimicry, which is crucial for heritage preservation and urban renewal. The proposed methodology integrates architectural typology with Low-Rank Adaptation (LoRA) for fine-tuning a Stable Diffusion (SD) model. This process involves a typology-guided preparation of a curated dataset (150 images) and precise control of training parameters. The resulting typologically guided LoRA-tuned model demonstrates significant performance improvements over baseline models. Quantitative analysis shows a 24.6% improvement in Fréchet Inception Distance (FID) and a 7.0% improvement in Learned Perceptual Image Patch Similarity (LPIPS). Furthermore, qualitative evaluations by 68 experts confirm superior realism and stylistic accuracy. The findings indicate that this synergy enables data-efficient, typology-grounded stylistic emulation, highlighting AI’s potential as a creative partner for nuanced reinterpretation. However, achieving deeper semantic understanding and robust 3D inference remains an ongoing challenge. Full article