Road potholes are a serious threat to road safety. They can damage vehicles and increase the risk of traffic accidents. However, factors like lighting and weather can impact how well computer vision systems detect potholes. To tackle this issue, this study compares four You Only Look Once (YOLO) 2026 (YOLO26) variants: YOLO26n, YOLO26s, YOLO26m, and YOLO26l, for detecting potholes in different environmental conditions. A dataset was independently collected in the Poris area of Tangerang, Indonesia, using a forward-facing camera to capture a driver's view. The video recordings were turned into annotated image frames in YOLO format, creating 368 images that included clear, rain, evening, and night conditions. Each model was trained for 200 epochs under the same experimental conditions. Their performance was evaluated using Precision, Recall, mean Average Precision at an Intersection over Union (IoU) threshold of 0.5 (mAP@0.5), and mean Average Precision across IoU thresholds from 0.5 to 0.95 (mAP@0.5:0.95). Computational efficiency was evaluated using parameter count, model size, training time, GPU memory usage, inference time, and frames per second (FPS) to assess the computational requirements and real-time detection capability of each YOLO26 variant. The results showed that YOLO26m achieving the highest Recall of 0.877 and mAP@0.5 of 0.934, while YOLO26l achieved the highest Precision of 0.989 and mAP@0.5:0.95 of 0.735. On the other hand, YOLO26n was the most computationally efficient option while still providing competitive detection performance. These results highlight the trade-off between detection accuracy and computational efficiency among YOLO26 variants. They also offer useful guidance for choosing the right model for real-time pothole detection in various conditions, which can help in the development of future driver assistance systems and pothole warning systems.
Domestic wastewater treatment plants (DWTPs) must keep pond effluent within Indonesia's mandated wastewater quality standards, yet most operators still rely on manual, edge-of-pond sampling, which is slow, labour-intensive, and cannot represent the whole pond surface. This study develops a solar-powered, remotely controlled mobile sensing platform integrated with Long Range (LoRa) communication and Internet of Things (IoT) connectivity for real-time domestic wastewater quality monitoring. The system comprises a Mobile System that floats on the pond, carrying temperature, pH, TDS, turbidity, and dissolved oxygen (DO) sensors with a solar panel, battery, and two DC-motor thrusters, and a Monitoring System at the pond edge that relays sensor data via LoRa to an LCD and the ThingSpeak platform, and issues four-directional motion commands to the Mobile System through a second LoRa link. The fabricated prototype was validated experimentally: against laboratory references (n = 30 per sensor), the five sensors achieved R² = 0.9654–0.9942 with MAPE below 5%; the dual LoRa links held packet loss below 4% up to 250 m line-of-sight; the ThingSpeak link achieved 98.92% data-delivery success with 1.35 s average latency over a 72-hour test; and the solar system supplied a daily surplus of 412 Wh against 152.7 Wh consumed, keeping battery charge above 81.5% throughout. This work contributes a mobile, solar-powered, LoRa- and IoT-integrated monitoring architecture, experimentally validated at prototype scale, enabling operators to obtain water quality data across the pond surface and redeploy the platform across multiple ponds without additional cabling.
Mobile application reviews provide critical unstructured feedback for developers, yet processing large-scale Indonesian app reviews remains hindered by informal language noise and severe class imbalance. While deep learning offers strong context modeling, its computational burden is impractical for industrial-scale deployment exceeding hundreds of thousands of entries. This study proposes a scalable, computationally efficient Hybrid Lexicon-Machine Learning framework for Aspect-Based Sentiment Analysis (ABSA) on large-scale Indonesian user reviews, using Telegram as a case study. We processed 399,527 reviews from the Google Play Store through a pipeline integrating TF-IDF n-grams, explicit domain-adapted lexicon scoring (incorporating negation and intensifier heuristics), and review length metadata into a LightGBM classifier. To overcome the scarcity and linguistic ambiguity of neutral ratings, a binary classification strategy (Positive vs. Negative/Critical) was implemented. The model achieved Accuracy of 80.4%, Macro-F1 of 76.7%, and AUC-ROC of 0.855 on a strictly held-out test set, peaking at 0.933 for Privacy. Ablation experiments confirm that incorporating explicit lexicon scores consistently improves Macro-F1 by up to 5.1% over standalone TF-IDF. Aspect distribution revealed critical bottlenecks in Privacy (73.0% negative) and Notification (66.1% negative), providing actionable roadmaps for software maintenance.
The Discrete Fourier Transform (DFT) allows the digital signals representation in the frequency domain by decomposing discrete signals into finite sinusoidal components. DFT works on discrete data with a certain number of samples relevant for modern digital signal processing. In the context of electroencephalogram (EEG), DFT plays an important role in identifying brain frequency spectra, such as alpha, beta, theta, and delta waves, each of which has physiological and clinical significance. This article analyzes the Discrete Fourier Transform (DFT) in the context of finite-dimensional Hilbert space and its application to analyze EEG signals. The objectives of the study are to prove that the DFT is in Hilbert space by showing its mathematical properties and using DFT to analyze the frequency characteristics of EEG signals in finite-dimensional Hilbert space. The method used involves a mathematical approach by establishing a mathematical framework based on finite-dimensional Hilbert space, analyzing the inverse DFT by recasting the signal from the frequency domain to the time domain, identifying the relevant mathematical characteristics of the DFT. Discretization of continuous signals into discrete ones through impulse sampling techniques, and application of DFT for transformation to the frequency domain for analyzing EEG spectral. The results show that DFT satisfies the definition and axioms of complex vector spaces and finite-dimensional Hilbert spaces, having a provable inverse through the properties of unitary operators. The application of DFT to the modeled EEG signal successfully produces a frequency spectrum that matches the frequency components of the initial model, both in magnitude and energy.
Tactical unmanned aerial vehicle (UAV) mesh networks must maintain useful communication under rapid topology changes, spectrum attack, and traffic pressure. This paper evaluates a Layer-2.5 Zero-Trust Adaptation Layer (ZTAL) as an in-path logical shim between IPv4/OLSR and IEEE 802.11p MAC/PHY. The final experiment uses 90 ns-3 simulations from final_matrix_v3 with Gauss-Markov mobility, 10 and 20 UAV, EDCA, REACTIVE_FH, and ZTAL treatments, and NONE, CW_JAMMING, and ALG_DOS conditions. At 10 UAV under the configured 50 dBm CW jammer, ZTAL improved packet delivery ratio from 57.71% with EDCA to 69.96% (paired difference +12.25 percentage points, Holm p=1.14e-5). At 20 UAV, ZTAL improved PDR from 26.83% to 45.95% (paired difference +19.13 percentage points, Holm p=6.04e-7). Relative to REACTIVE_FH, ZTAL showed higher mean PDR and more reliable coordination, but the PDR advantage was not statistically conclusive after multiplicity correction with n=5. The results also show no-attack overhead, lower delivered byte throughput in several cases, and substantial queue expiry under high-density/load stress. A separate Hard Purge microbenchmark reduced peak protocol-managed reassembly payload state from 192,000 B to 96,020 B under the tested authenticated overlap pattern. The findings support bounded simulation evidence for CW resilience and state-release behavior, not field deployment readiness.
