Search Results (47)

While objective street metrics derived from imagery or GIS have become standard in urban analytics, they remain insufficient to capture subjective perceptions essential to inclusive urban design. This study introduces a novel Multimodal Street Evaluation Framework (MSEF) that fuses a vision transformer (VisualGLM-6B) with a large language model (GPT-4), enabling interpretable dual-output assessment of streetscapes. Leveraging over 15,000 annotated street-view images from Harbin, China, we fine-tune the framework using Low-Rank Adaptation(LoRA) and P-Tuning v2 for parameter-efficient adaptation. The model achieves an F1 score of 0.863 on objective features and 89.3% agreement with aggregated resident perceptions, validated across stratified socioeconomic geographies. Beyond classification accuracy, MSEF captures context-dependent contradictions: for instance, informal commerce boosts perceived vibrancy while simultaneously reducing pedestrian comfort. It also identifies nonlinear and semantically contingent patterns—such as the divergent perceptual effects of architectural transparency across residential and commercial zones—revealing the limits of universal spatial heuristics. By generating natural-language rationales grounded in attention mechanisms, the framework bridges sensory data with socio-affective inference, enabling transparent diagnostics aligned with Sustainable Development Goal 11(SDG 11). This work offers both methodological innovation in urban perception modeling and practical utility for planning systems seeking to reconcile infrastructural precision with lived experience. Full article

The deployment of transformer-based language models on resource-constrained edge devices presents fundamental challenges in computational efficiency and memory utilization. We introduce SQ-LoRA (Stable-rank Quantized Low-Rank Adaptation), a theoretically grounded compression framework that achieves unprecedented efficiency through the synergistic integration of adaptive low-rank decomposition, hardware-accelerated structured sparsity, and intelligent hybrid quantization. Our primary contribution establishes the first rigorous mathematical connection between the matrix stable rank and optimal LoRA rank selection, formalized in Theorem I, which provides bounded approximation guarantees. SQ-LoRA implements: (1) adaptive rank allocation via stable-rank analysis to automatically determine layer-wise compression ratios; (2) 4:8 structured sparsity patterns, enabling 2× hardware acceleration on modern edge processors; and (3) a three-tier quantization scheme that combines 4-bit NormalFloat storage with selective 3-bit/8-bit precision to preserve outliers. A comprehensive evaluation on four diverse natural language processing (NLP) benchmarks demonstrates that SQ-LoRA achieves a 320 MB memory footprint (96.7% reduction) and a 10 ms inference latency (91.7% improvement), and maintains 82.0% average accuracy (within 0.15% of the full model). Statistical significance testing (p < 0.001) confirms its superiority over state-of-the-art methods. This framework enables the deployment of sophisticated language models on devices with 2 GB of RAM, advancing practical edge-AI applications. Full article

As an important ecological obstacle prone to climatic changes, the Tibetan Plateau has been transformed by retreating glaciers, degrading permafrost, and deteriorating grasslands. Recent ecological remote sensing evaluations typically use medium-resolution and single-source optical imagery, highlight natural factors while ignoring human impacts, and encounter difficulties with time-focused interpretability and continuity within complex terrains. This research proposes a theory combining large language models with satellite embedding to holistically examine the ecology of the Tibetan Plateau between 2000 and 2024. We created an ecological satellite embedding (ESE) model applying self-supervised learning to integrate 12 ecological variables into combined space and time representations as of 2024, according to the Prithvi-Earth Observation (Prithvi-EO) foundational model involving low-rank adaptation (LoRA). GeoChat reasoning was applied to turn the embedded variables into a comprehensive representation feature (CRF). Field research demonstrated strong accuracy for the fraction of absorbed photosynthetically active radiation (FAPAR, R² = 0.9923) and aboveground biomass (AGB, R² = 0.8690). Space and temporal analyses demonstrated a general ecology-dependent enhancement accompanied by significant space-based clustering (Moran’s I = 0.50–0.80), hotspots in humid southeastern areas, major upward trends in vegetation indices and productivity metrics (p < 0.05), and higher shifts in transition regions. Despite the marginal degradation risk, the grassland carrying capacity has expanded extensively in the main farming regions. The comprehensible CRF schema identified three management areas: potential risk, enhancement potential, and stable conservation management. This transferable modular approach connects expert reasoning with data-driven modeling, presenting adaptable methods for assessing ecosystems in high-altitude, data-sparse environments, and practical ways to promote ecological management. Full article

In this work, the authors introduce an entirely solar-powered LoRa-based WSN consisting of several nodes, two stoplights, and four cameras. The system has been used to monitor the semi-rural area of Panni (FG), Puglia, Italy. The WSN has a totally custom implementation in both the node-gateway side and the gateway-user interface side. In particular, the communication framework is entirely IoT-based, featuring both the MQTT protocol, for the direct control of apparatuses from the system user interface, and the more traditional TCP/IP protocol, implemented on NB-IoT. The proposed system is entirely solar-powered and features a 34.68 mWh/day consumption. Around a single communication session, the average power consumption inside the single node amounts to 1.4 mW. This paper gives an overview of the proposed system, with detailed explanations of each part, and measurements retrieved over a wide period to assess the functionality of the system. Full article

This paper examines precise timestamping of LoRaWAN messages (particularly beacons) to enable wide-area synchronization for end devices without GNSS. The need for accuracy demands hardware-level timestamping architectures, possibly using time-domain cross-correlation (matched filtering) against internally generated chirp references. Focusing on Time-of-Arrival (TOA) estimation from raw IQ samples, the authors analyze effects of non-idealities—additive white Gaussian noise (AWGN), Carrier Frequency Offset (CFO), Sampling Phase and Frequency Offset (SPO and SFO, respectively), and radio parameters such as spreading factor (SF) and sampling rate of the baseband signals. A MATLAB (R2020) simulation mimics preamble detection and Start-of-Frame Delimiter (SFD) timestamping while sweeping SF (7, 9, 12), sampling rates (0.25–10 MSa/s), SNR (−20 to +20 dB), and CFO/SFO offsets (−10–10 ppm frequency deviation). Errors are evaluated in terms of mean and dispersion, the latter represented by the P95–P5 range metric. Results show that oversampling not only improves temporal resolution, but sub-microsecond error dispersion can be achieved with high sampling rates in favorable SNR and SF cases. Indeed, SPO and SNR greatly contribute to error dispersion. On the other hand, higher SF values increase correlation robustness at the cost of longer chirps, making SFO a dominant error source; ±10 ppm SFO can induce roughly ±3 μs SFD bias for SF12. CFO largely cancels after up-/down-chirp averaging. As a concluding remark, matched-filter hardware timestamping can ensure sub-μs errors thanks to oversampling but requires SFO compensation for accurate real-world synchronization in practice. Full article

This study aims to investigate the predictive value of pre-treatment multi-phasic contrast-enhanced computed tomography (CECT) radiomic features for treatment resistance in patients with rat sarcoma virus (RAS)-mutated colorectal liver metastases (CRLMs) receiving bevacizumab-based chemotherapy. Seventy-three samples with RAS-mutated CRLMs receiving bevacizumab-combined chemotherapy regimens were evaluated. Radiomic features were extracted from arterial phase (AP), portal venous phase (PVP), AP-PVP subtraction image, and Delta phase (DeltaP, calculated as AP-to-PVP ratio) images. Three groups of radiomics features were extracted for each phase, including peritumor, core tumor, and whole-tumor regions. For each of the four phases, a two-sided independent Mann–Whitney U test with the Bonferroni correction and K-means clustering was applied to the remnant features for each phase. Subsequently, the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was then applied for further feature selection. Six machine learning algorithms were then used for model development and validated on the independent testing cohort. Results showed peritumoral radiomic features and features derived from Laplacian of Gaussian (LoG) filtered images were dominant in all the compared machine learning algorithms; NB models yielded the best-performing prediction (Avg. training AUC: 0.731, Avg. testing AUC: 0.717) when combining all features from different phases of CECT images. This study demonstrates that peritumoral radiomic features and LoG-filtered pre-treatment multi-phasic CECT images were more predictive of treatment response to bevacizumab-based chemotherapy in RAS-mutated CRLMs compared to core tumor features. Full article

The advancement of Automotive Industry 4.0 has promoted the development of Vehicle to Vehicle (V2V) and Internet of Vehicles (IoV) communication, which marks the new era for intelligent, connected and automated transportation. Despite the benefits of this metamorphosis in terms of effectiveness and convenience, new obstacles to safety, inter-connectivity, and cybersecurity emerge. The tire pressure monitoring system (TPMS) is one prominent feature that senses tire pressure, which is closely related to vehicle stability, braking performance and fuel efficiency. However, the majority of TPMSs currently in use are based on the use of insecure and proprietary wireless communication links that can be breached by attackers so as to interfere with not only tire pressure readings but also sensor data manipulation. For this purpose, we design a secure TPMS architecture suitable for real-time IoV sensing. The framework is experimentally implemented using a Raspberry Pi 3B+ (Raspberry Pi Ltd., Cambridge, UK) as an independent autonomous control unit (ACU), interfaced with vehicular pressure sensors and a LoRa SX1278 (Semtech Corporation, Camarillo, CA, USA) module to support low-power, long-range communication. The gathered sensor data are encrypted, their integrity checked, source authenticated by lightweight cryptographic algorithms and sent to a secure server locally. To validate this approach, we show a three-node exhibition where Node A (raw data and tampered copy), B (unprotected copy) and C (secure auditor equipped with alerting of tampering and weekly rotation of the ID) realize detection of physical level threats at top speeds. The validated datasets are further enriched in a MATLAB R2024a simulator by replicating the data of one vehicle by 100 virtual vehicles communicating using over 5G, LoRaWAN and LoRa P2P as communication protocols under urban, rural and hill-station scenarios. The presented statistics show that, despite 5G ultra-low latency, LoRa P2P consistently provides better reliability and energy efficiency and is more resistant to attacks in the presence of various terrains. Considering the lack of private vehicular 5G infrastructure and the regulatory restrictions, this work simulated and evaluated the performance of 5G communication, while LoRa-based communication was experimentally validated with a hardware prototype. The results underline the trade-offs among LoRa P2P and an infrastructure-based uplink 5G mode, when under some specific simulation conditions, as opposed to claiming superiority over all 5G modes. In conclusion, the presented Raspberry Pi–MATLAB hybrid solution proves to be an effective and scalable approach to secure TPMS in IoV settings, intersecting real-world sensing with large-scale network simulation, thus enabling safer and smarter next-generation vehicular systems. Full article

This study aims to establish a method-integrative framework for emotion-oriented architectural image generation. The framework combines Stable Diffusion with targeted LoRA (Low-Rank Adaptation), a lightweight and parameter-efficient fine-tuning approach, together with ControlNet-based structural constraints, to examine how controllable design-element manipulations influence emotional responses. The methodology follows a closed-loop “generation–evaluation” workflow, with each LoRA module independently targeting a single design element. Guided by the relaxation–arousal emotional dimension, the framework is evaluated using subjective ratings and electroencephalogram (EEG) measures. Twenty-seven participants viewed six architectural space categories, each comprising four conditions (baseline, color, material, and form modification). EEG α/β power ratio (RAB) served as the primary neurophysiological marker of arousal. Statistical analysis indicated that LoRA-based modifications of design elements produced distinct emotional responses: color and material changes induced lower arousal, whereas changes in form elicited a bidirectional pattern involving relaxation and arousal. The right parietal P4 electrode site showed the most sensitive emotional response to design element changes, with consistent statistical significance. P4 is a human scalp EEG location associated with cortical activity related to visuospatial processing. Descriptive results suggested opposite directional effects with similar intensity trends; however, linear mixed-effects model (LMM) inference did not support significant group-level linear coupling, indicating individual variation. This study demonstrates the feasibility of emotion-guided architectural image generation, showing that controlled manipulation of color, material, and form can elicit measurable emotional responses in human brain activity. The findings provide a methodological basis for future multimodal, adaptive generative systems and offer a quantitative pathway for investigating the relationship between emotional states and architectural design elements. Full article

Tracking small, fast-moving unmanned aerial vehicles (UAVs) in thermal infrared (TIR) imagery is a significant challenge due to low-resolution targets, Dynamic Background Clutter, and frequent occlusions. To address this, we introduce MemLoTrack, a novel onestream Vision Transformer tracker that integrates a memory mechanism into a parameterefficient LoRA framework. MemLoTrack enhances a baseline tracker (LoRAT) with two key components: (i) a gated First-In, First-Out (FIFO) memory bank (MB) for temporal context aggregation and (ii) a lightweight Memory Attention Layer (MAL) for effective information retrieval. A key component of our method is a selective memory update policy, which commits a frame to the memory bank only when it satisfies both a classification confidence threshold ($\tau$) and a Kalman filter-based motion consistency check. This gating mechanism robustly prevents memory contamination due to distractors, occlusions, and reappearance events. Our training is highly efficient, updating only the LoRA adapters, MAL, and prediction head while the pretrained DINOv2 backbone remains frozen. Evaluated on the challenging Anti-UAV410 benchmark, MemLoTrack (L_mem = 7, $\tau$ = 0.8) achieves an AUC of 63.6 and a State Accuracy (SA) of 64.0, representing a significant improvement over the LoRAT baseline by +1.4 AUC and +1.5 SA. Compared to the state-of-the-art method FocusTrack, MemLoTrack demonstrates superior robustness with higher AUC (63.6 vs. 62.8) and SA (64.0 vs. 63.9), while trading lower precision (P/P-Norm) scores. Furthermore, MemLoTrack operates at 153 FPS on a single RTX 4070 Ti SUPER, demonstrating that parameter-efficient fine-tuning with a selective memory mechanism is a powerful and deployable strategy for real-time Anti-UAV tracking in demanding TIR environments. Full article

The Green Internet of Things (GIoTs) has emerged as a transformative paradigm for environmental conservation, enabling autonomous, self-sustaining sensor networks that operate without batteries and with minimal ecological footprint. This approach is especially critical for long-term mangrove monitoring in smart coastal cities, where conventional battery-powered systems are impractical due to frequent, costly, and environmentally disruptive replacements that hinder continuous data collection. This paper presents a self-sustaining GIoT sensing system for mangrove monitoring powered by sedimentary microbial fuel cells (SMFCs), enabling perpetual, battery-less, and zero-emission operation. A spatial dynamic energy management (DPM) strategy is implemented for the efficient integration of a microcontroller unit with a LoRa wireless communication transceiver and the SMFC harvested energy, ensuring a balanced self-sustained approach into a GIoT sensing network. Experimental results demonstrate an average power consumption of 190.45 μW per 14-byte data packet transmission, with each packet containing pH, electrical conductivity and ambient temperature measurements from the mangrove environment. Under the spatial DPM strategy, the network of four sensing nodes exhibited an energy consumption of 1.14 mWh. Given a harvested power density of 15.1 mW/m² from the SMFC, and utilizing a 0.1 F supercapacitor as an energy buffer, the system can support at least six consecutive data transmissions. These findings validate the feasibility of sustainable, low-power GIoT architectures for ecological monitoring. Full article

To enable remote and automatic monitoring of the farmland soil information, this paper has developed a soil monitoring system based on the Internet of Things (IoT), which mainly involves the development of a gateway server node, wireless sensor nodes, a remote monitoring platform, and photovoltaic (PV) modules. The Raspberry Pi 5-based gateway server periodically sends data acquisition commands to wireless sensor nodes via LoRa, receives soil data returned by sensor nodes, and stores them in a MySQL database. Using a remote monitoring platform, Internet users can monitor real-time and historical soil data stored in the database. The STM32F103C8T6-based wireless sensor node receives data acquisition commands from the gateway server, uses soil temperature and humidity sensors as well as a pH sensor to collect soil status, and then sends sensor data back to the gateway server via LoRa. The system is powered by both PV energy and batteries, which enhances the endurance capability. Experimental results show that the designed system works well in remotely monitoring soil information. Using the proposed query attempt dynamic adjustment (QADA) method, the wireless sensor node dynamically adjusts the number of query attempts, which reduces the data acquisition failure rate from 21–25% to no more than 0.33%. Using the obtained qualitative relationship that the data acquisition delay varies inversely with the LoRa transfer rate, the data acquisition delay can be reduced to less than 67 ms. Full article

Subsoil compaction is a major constraint in sugarcane production, limiting yields and reducing resource-use efficiency. This study presents WireDepth, an innovative cloud-connected monitoring system that leverages edge computing and IoT technologies for real-time, spatially aware analysis and visualization of subsoiling depth. The system integrates ultrasonic, laser, inclinometer, and potentiometer sensors mounted on the subsoiler, with on-board microcontroller processing and dual wireless connectivity (LoRaWAN and NB-IoT/LTE-M) for robust data transmission. A cloud platform delivers advanced analytics, including 3D depth maps and operational efficiency metrics. System accuracy was assessed using 300 reference depth measurements, with Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) calculated per sensor. The inclinometer and potentiometer achieved the highest accuracy (MAPE of 0.92% and 0.84%, respectively), with no significant deviation from field measurements (paired t-tests, p > 0.05). Ultrasonic and laser sensors exhibited higher errors, particularly at shallow depths, due to soil debris interference. Correlation analysis confirmed a significant effect of depth on sensor accuracy, with laser sensors showing the strongest association (Pearson r = 0.457, p < 0.001). Field validation in commercial sugarcane fields demonstrated that WireDepth improves subsoiling precision, reduces energy waste, and supports sustainable production by enhancing soil structure and root development. These findings advance precision agriculture by offering a scalable, real-time solution for subsoiling management, with broad implications for yield improvement in compaction-affected systems. Full article

The growing demand for electricity and the urgent need to reduce environmental impact have made sustainable energy utilization a global priority. Street lighting, as a significant consumer of urban electricity, requires innovative solutions to enhance efficiency and reliability. This study presents an off-grid smart street lighting system that combines solar photovoltaic generation with battery storage and Internet of Things (IoT)-based control to ensure continuous and efficient operation. The system integrates Long Range Wide Area Network (LoRaWAN) communication technology for remote monitoring and control without internet connectivity and employs the Perturb and Observe (P&O) maximum power point tracking (MPPT) algorithm to maximize energy extraction from solar sources. Data transmission from the LoRaWAN gateway to the cloud is facilitated through the Message Queuing Telemetry Transport (MQTT) protocol, enabling real-time access and management via a graphical user interface. Experimental results demonstrate that the proposed system achieves a maximum MPPT efficiency of 97.96%, supports reliable communication over distances of up to 10 km, and successfully operates four LED streetlights, each spaced 400 m apart, across an open area of approximately 1.2 km—delivering a practical, energy-efficient, and internet-independent solution for smart urban infrastructure. Full article

In the rapidly evolving landscape of digital marketing and electronic commerce, short-form content—particularly on platforms like Twitter (now X)—has become pivotal for real-time branding, community engagement, and product promotion. The rise of Non-Fungible Tokens (NFTs) and Web3 ecosystems further underscores the need for domain-specific, engagement-oriented social media content. However, automating the generation of such content while balancing linguistic quality, semantic relevance, and audience engagement remains a substantial challenge. To address this, we propose RL-TweetGen, a socio-technical framework that integrates instruction-tuned large language models (LLMs) with reinforcement learning (RL) to generate concise, impactful, and engagement-optimized tweets. The framework incorporates a structured pipeline comprising domain-specific data curation, semantic classification, and intent-aware prompt engineering, and leverages Parameter-Efficient Fine-Tuning (PEFT) with LoRA for scalable model adaptation. We fine-tuned and evaluated three LLMs—LLaMA-3.1-8B, Mistral-7B Instruct, and DeepSeek 7B Chat—guided by a hybrid reward function that blends XGBoost-predicted engagement scores with expert-in-the-loop feedback. To enhance lexical diversity and contextual alignment, we implemented advanced decoding strategies, including Tailored Beam Search, Enhanced Top-p Sampling, and Contextual Temperature Scaling. A case study focused on NFT-related tweet generation demonstrated the practical effectiveness of RL-TweetGen. Experimental results showed that Mistral-7B achieved the highest lexical fluency (BLEU: 0.2285), LLaMA-3.1 exhibited superior semantic precision (BERT-F1: 0.8155), while DeepSeek 7B provided balanced performance. Overall, RL-TweetGen presents a scalable and adaptive solution for marketers, content strategists, and Web3 platforms seeking to automate and optimize social media engagement. The framework advances the role of generative AI in digital commerce by aligning content generation with platform dynamics, user preferences, and marketing goals. Full article

Retrieval-Augmented Generation (RAG) combined with Large Language Models (LLMs) introduces a new paradigm for clinical-trial data analysis that is both real-time and knowledge-traceable. This study targets a multi-site, real-world data environment. It builds a hierarchical RAG pipeline spanning an electronic health record (EHR), National Health Insurance (NHI) billing codes, and image-vector indices. The LLM is optimized through lightweight LoRA/QLoRA fine-tuning and reinforcement-learning-based alignment. The system first retrieves key textual and imaging evidence from heterogeneous data repositories and then fuses these artifacts into the contextual window for clinical report generation. Experimental results show marked improvements over traditional manual statistics and prompt-only models in retrieval accuracy, textual coherence, and response latency while reducing human error and workload. In evaluation, the proposed multimodal RAG-LLM workflow achieved statistically significant gains in three core metrics—recall, factual consistency, and expert ratings—and substantially shortened overall report-generation time, demonstrating clear efficiency advantages versus conventional manual processes. However, LLMs alone often face challenges such as limited real-world grounding, hallucination risks, and restricted context windows. Similarly, RAG systems, while improving factual consistency, depend heavily on retrieval quality and may yield incoherent synthesis if evidence is misaligned. These limitations underline the complementary nature of integrating RAG and LLM architectures in a clinical reporting context. Quantitatively, the proposed system achieved a Composite Quality Index (CQI) of 78.3, outperforming strong baselines such as Med-PaLM 2 (72.6) and PMC-LLaMA (74.3), and reducing the report drafting time by over 75% (p < 0.01). These findings confirm the practical feasibility of the framework to support fully automated clinical reporting. Full article