Understanding the quality of life related to transportation plays a crucial role in enhancing commuters’ life, particularly daily trips. This study explores spatial effects built environment on (QoLT) through combination GIS application and deep learning based questionnaire survey by focusing case Sukhumvit district, Bangkok, Thailand. The Geographic Information System (GIS) was applied for analysis visualization among all variables grid cell (500 × 500 sq.m.). In regard learning, semantic...

The precise classification of crop types using hyperspectral remote sensing imaging is an essential application in the field agriculture, and significance for yield estimation growth monitoring. Among deep learning methods, Convolutional Neural Networks (CNNs) are premier model image (HSI) their outstanding locally contextual modeling capability, which facilitates spatial spectral feature extraction. Nevertheless, existing CNNs have a fixed shape limited to observing restricted receptive...

(1) Computer Vision: The field of computer vision is making significant strides in dynamic reasoning capability through test-time scaling (TTS) [...]

In recent years, methods based on deep convolutional neural networks (CNNs) have dominated the classification task of hyperspectral images. Although CNN-based HSI advantages spatial feature extraction, images are characterized by approximately continuous spectral information, usually containing hundreds bands. CNN cannot mine and represent sequence properties features well, transformer model attention mechanism proves its in processing data. This study proposes a new kernel combined with...

Traditional endoscopic treatment methods restrict the surgeon’s field of view. New approaches to laparoscopic visualization have emerged due advent robot-assisted surgical techniques. Lumen simultaneous localization and mapping (SLAM) technology can use image sequence taken by endoscope estimate pose reconstruct lumen scene in minimally invasive surgery. This gives surgeon better visual perception is basis for development navigation systems as well medical augmented reality. However,...

Although the collaborative use of hyperspectral images (HSIs) and LiDAR data in land cover classification tasks has demonstrated significant importance potential, several challenges remain. Notably, heterogeneity cross-modal information integration presents a major obstacle. Furthermore, most existing research relies heavily on category names, neglecting rich contextual from language descriptions. Visual-language pretraining (VLP) achieved notable success image recognition within natural...

Recently, deep learning techniques, specifically semantic segmentation, have been employed to extract visual features from street images, a dimension that has received limited attention in the investigation of connection between subjective and objective road environment perception. This study is dedicated exploring comprehending factors influencing commuters’ perceptions environment, with aim bridging gap interpreting environmental quality Thailand. Semantic segmentation was applied identify...

Aiming to solve the problems of different spectral bands and spatial pixels contributing differently hyperspectral image (HSI) classification, sparse connectivity restricting convolutional neural network a globally dependent capture, we propose HSI classification model combined with multi-scale residual spectral–spatial attention an improved transformer in this paper. First, order efficiently highlight discriminative information, feature extraction module that preserves information two-layer...

X-ray security images face significant challenges due to complex backgrounds, item overlap, and multi-scale target detection. Traditional methods often struggle accurately identify objects, especially under cluttered conditions. This paper presents an advanced detection model, called YOLOv8n-GEMA, which incorporates several enhancements address these issues. Firstly, the generalized efficient layer aggregation network (GELAN) module is employed augment feature fusion capabilities. Secondly,...

Deep-learning-based multi-sensor hyperspectral image classification algorithms can automatically acquire the advanced features of multiple sensor images, enabling model to better characterize data and improve accuracy. However, currently available methods for feature representation in remote sensing their respective domains do not focus on existence bottlenecks heterogeneous fusion due different sensors. This problem directly limits final collaborative performance. In this paper, address...

Convolutional neural networks (CNNs) are indeed commonly employed for hyperspectral image classification. However, the architecture of cellular typically requires manual design and fine-tuning, which can be quite laborious. Fortunately, there have been recent advancements in field Neural Architecture Search (NAS) that enable automatic networks. These NAS techniques significantly improved accuracy HSI classification, pushing it to new levels. This article proposes a Multi-Scale...

The depth information of abdominal tissue surface and the position laparoscope are very important for accurate surgical navigation in computer-aided surgery. It is difficult to determine lesion location by empirically matching laparoscopic visual field with preoperative image, which easy cause intraoperative errors. Aiming at complex environment, this paper constructs an improved monocular simultaneous localization mapping (SLAM) system model, can more accurately truly reflect cavity...

In view of the complexity and diversity hyperspectral images (HSIs), classification task has been a major challenge in field remote sensing image processing. Hyperspectral (HSIC) methods based on neural architecture search (NAS) is current attractive frontier that not only automatically searches for network architectures best suited to characteristics HSI data, but also avoids possible limitations manual design networks when dealing with new tasks. However, existing NAS-based HSIC have...

Cross-project software defect prediction (CPDP) refers to the construction of models by collecting multi-source project data, but heterogeneity data among projects and modern problem “data islands” hinder its development. In response these challenges, we propose a CPDP algorithm based on differential perception combined with inheritance federated learning (FedDPI). Firstly, design an efficient preprocessing scheme, which lays reliable foundation for integrating oversampling optimal feature...

For an urban development, the Quality of Life (QOL) people in city is a vital issue that should be considered. There are many researches QOL topics use questionnaire survey approach. These studies yield very useful information for development planning. As Artificial Intelligence technologies developed fast recently, they applied to solve transportation problems. In this paper, we propose method automatically extract mobility indicators using two image recognition techniques: Semantic...

Convolution neural network (CNN)is widely used in hyperspectral image (HSI) classification. However, the architecture of CNNs is usually designed manually, which requires careful fine-tuning. Recently, many technologies for search (NAS) have been proposed to automatically design networks, further improving accuracy HSI classification a new level. This paper proposes circular kernel convolution- β -decay regulation NAS-confident learning rate (CK- NAS-CLR) framework structure First, this...

The well-being of residents is a top priority for megacities, which why urban design and sustainable development are crucial topics. Quality Life (QoL) used as an effective key performance index (KPI) to measure the efficiency city plan’s quantity quality factors. For dwellers, QoL pedestrians also significant. walkability concept evaluates analyzes in walking scene. However, traditional questionnaire survey approach costly, time-consuming, limited its evaluation area. To overcome these...

With the increase of population aging, chronic diseases and accidental injuries, more people are facing plight diminished or even lost walking ability. As a kind mobile service robot, smart wheelchair has strong environmental adaptability, smooth motion control friendly human-computer interaction experience, is an indispensable tool in rehabilitation engineering elderly assistance engineering, which important research value social significance. In this paper, system structure framework...

To address the issue of reduced gaze estimation accuracy caused by individual differences in different environments, this study proposes a novel algorithm based on attention mechanisms. Firstly, constructing facial feature extractor (FFE), method obtains information about eyes and locates areas left right eyes. Then, L2CSNet (l2 loss + cross-entropy softmax layer network), which integrates PSA (pyramid squeeze attention), is designed to increase correlation weights related areas, suppress...

Hyperspectral imagery (HSI) and light detection ranging (LiDAR) remote sensing technologies are important ways to obtain surface information. Combining the characteristics advantages of hyperspectral images LiDAR DSM data, we propose a classification method for co-classification data in this paper. The model uses dual-branch HSI image method. First, features shallow convolution deep merged spatial feature extraction process HSI, which combines focus on more global Then space spectrum...

How to obtain internal cavity features and perform image matching is a great challenge for laparoscopic 3D reconstruction. This paper proposes method detecting associating vascular based on dual-branch weighted fusion structure enhancement. Our proposed divided into three stages, including analyzing various types of minimally invasive surgery (MIS) images designing universal preprocessing framework make our generalized. We propose Gaussian enhancement algorithm using the Frangi measure MFAT...

针对高光谱图像(hyperspectral images, HSI)与LiDAR数据多模态分类任务中的跨模态信息表达和特征对齐等问题,提出一种基于对比学习CNN-Transformer高光谱和LiDAR数据协同分类网络(Contrastive Learning based CNN-Transformer Network, CLCT-Net)。CLCT-Net通过由ConvNeXt V2 Block构成的共有特征提取模块,获得不同模态间的共性特征,解决异构传感器数据之间语义对齐的问题。构建了包含空间-通道分支和光谱上下文分支的双分支HSI编码器,以及结合频域自注意力机制的LiDAR编码器,以获取更丰富的特征表示。利用集成对比学习进行分类,进一步提升多模态数据协同分类的精度。在 Houston 2013 和 Trento 数据集上的实验结果表明,相较于其他高光谱图像和LiDAR数据分类模型,本文所提模型获得了更高的地物分类精度,分别达到了92.01%和98.90%,实现了跨模态数据特征的深度挖掘和协同提取。