Accelerating Solar PV Site Selection: YOLO-Based Identification of Sound Barriers Along Highways

Energies, Год журнала: 2025, Номер 18(9), С. 2366 - 2366

Опубликована: Май 6, 2025

The exponential growth of the installation solar photovoltaic systems has been a significant step in energy transition toward reducing dependence on fossil fuels and mitigating climate change. This raised concerns about land use, particularly regions where large tracts are allocated to farms. Highway infrastructures such as sound barriers occupy surfaces which under-utilized could therefore contribute renewable generation without increasing use. study proposes application YOLO object detection algorithm automatically identify analyse locations along highways using video or image data, estimate potential output from installed these barriers. model trained tested Portuguese highways, achieving mean average precision exceeding 0.84 for YOLOv10 when training datasets containing more than 600 images. Using geolocation images identification number YOLO, it is possible electricity inform decision makers technical–economic feasibility this infrastructure generation.

Язык: Английский

---

Computational Analysis Method for Multi-subject Behavior in Public Spaces Based on Targeted Computer Vision Tracking

Landscape Architecture, Год журнала: 2025, Номер 32(5), С. 29 - 36

Опубликована: Май 1, 2025

Язык: Английский

---

Efficient Transformer-Based Road Scene Segmentation Approach with Attention-Guided Decoding for Memory-Constrained Systems

Machines, Год журнала: 2025, Номер 13(6), С. 466 - 466

Опубликована: Май 28, 2025

Accurate object detection and an understanding of the surroundings are key requirements when applying computer vision systems in automotive or robotics industries, namely with autonomous vehicles self-driving robots. A precise road users obstacles is essential to avoid potential accidents. Due presence many objects diversity environment, segmentation scene remains a challenging task. In our approach, Transformer-based backbone employed for robust feature extraction encoder module. addition, we have developed custom decoder module which implement attention-based fusion mechanisms effectively combine features. The modification specifically designed maintain fine spatial details enhance global context understanding, setting method apart from conventional approaches that typically use simple projection layers standard query-based decoders. implemented model consists 17.2 million parameters achieves competitive performance, mean intersection over union (mIoU) 76.41% on Cityscapes validation set. results gathered indicate ability capture both critical accurate urban scenes. Furthermore, lightweight design makes approach suitable deployment memory-limited devices.

Язык: Английский

---

LighterFace Model for Community Face Detection and Recognition

Information, Год журнала: 2024, Номер 15(4), С. 215 - 215

Опубликована: Апрель 11, 2024

This research proposes a face detection algorithm named LighterFace, which is aimed at enhancing speed to meet the demands of real-time community applications. Two pre-trained convolutional neural networks are combined, namely Cross Stage Partial Network (CSPNet), and ShuffleNetv2. Connecting optimized network with Global Attention Mechanism (GAMAttention) extends model compensate for accuracy loss caused by optimizing structure. Additionally, learning rate dynamically updated using cosine annealing method, enhances convergence during training. paper analyzes training LighterFace on WiderFace dataset custom dataset, aiming classify faces in real-life settings. Compared mainstream YOLOv5 model, demonstrates significant reduction computational 85.4% while achieving 66.3% increase attaining 90.6% detection. It worth noting that generates high-quality cropped images, providing valuable inputs subsequent recognition models such as DeepID. specifically designed run edge devices lower capabilities. Its performance Raspberry Pi 3B+ validates results.

Язык: Английский

---

Unlocking AI Adoption in Local Governments: Best Practice Lessons from Smart Cities

Опубликована: Июнь 4, 2024

In an era marked by swift technological progress, the pivotal role of Artificial Intelligence (AI) is increasingly evident across various sectors, including local governments. These governmental bodies are progressively leveraging AI technologies to enhance service delivery their communities, ranging from simple task automation more complicated engineering endeavours. While and governments adopting AI, it imperative understand functions, implications, consequences AI. Despite growing importance this domain, a significant gap persists within scholarly discourse. This study strives bridge void exploring applications context government provision using inquiry generate lessons best practices for similar smart city initiatives. Through comprehensive grey literature review, we analysed 262 real-world implementations 170 worldwide. The findings underscore several key points: (a) There has been consistent upward trajectory in adoption over last decade; (b) Local China, US, UK at forefront adoption; (c) Among technologies, Natural Language Processing Robotic Process Automation emerge as most prevalent ones; (d) primarily deploy 28 distinct services; (e) Information management, back-office work, transportation traffic management leading domains terms adoption. enriches extant body knowledge providing overview existing sphere governance. It offers insights policymakers decision-makers considering adoption, expansion, or refinement urban provision. Additionally, underscores these guide successful integration optimisation future projects, ensuring they meet evolving needs communities.

Язык: Английский

---