Research on Park Perception and Understanding Methods Based on Multimodal Text–Image Data and Bidirectional Attention Mechanism

Buildings, Год журнала: 2025, Номер 15(9), С. 1552 - 1552

Опубликована: Май 4, 2025

Parks are an important component of urban ecosystems, yet traditional research often relies on single-modal data, such as text or images alone, making it difficult to comprehensively and accurately capture the complex emotional experiences visitors their relationships with environment. This study proposes a park perception understanding model based multimodal text–image data bidirectional attention mechanism. By integrating image incorporates encoder representations from transformers (BERT)-based feature extraction module, Swin Transformer-based cross-attention fusion enabling more precise assessment visitors’ in parks. Experimental results show that compared methods residual network (ResNet), recurrent neural (RNN), long short-term memory (LSTM), proposed achieves significant advantages across multiple evaluation metrics, including mean squared error (MSE), absolute (MAE), root (RMSE), coefficient determination (R2). Furthermore, using SHapley Additive exPlanations (SHAP) method, this identified key factors influencing experiences, “water”, “green”, “sky”, providing scientific basis for management optimization.

Язык: Английский

---

Leveraging generative AI synthetic and social media data for content generalizability to overcome data constraints in vision deep learning

Artificial Intelligence Review, Год журнала: 2025, Номер 58(5)

Опубликована: Фев. 24, 2025

Язык: Английский

---

Cross Border E-commerce Intelligent Image Recognition and Processing System Based on Deep Learning

Learning and analytics in intelligent systems, Год журнала: 2025, Номер unknown, С. 186 - 194

Опубликована: Янв. 1, 2025

Язык: Английский

---

Big Data in the Internet of Things IoT Sensor Data Analysis and Edge Computing

Advances in computational intelligence and robotics book series, Год журнала: 2025, Номер unknown, С. 79 - 94

Опубликована: Апрель 17, 2025

The rapid technological advancements have given machines the power to think and decide. This has been largely fueled by developments in 5G, IoT, Big Data, Cloud, Edge Computing, AI technologies such as Machine Learning, Deep Computer Vision. chapter on IoT discusses exponential growth is edge computing, which makes data processing faster secure compared a traditional cloud system. It identifies some computing applications AWS that involve Data architectures with low-cost techniques for denoising sensor fusion achieve actionables. takes its audience through integration of technologies, facilitated an architecture diagram, applied smart buildings, vehicles, traffic management, factories, surveillance systems, discussing challenges privacy security avenues will be open innovations future research opportunities within IoT-enabled systems.

Язык: Английский

---

Semantic enhancement and cross-modal interaction fusion for sentiment analysis in social media

PLoS ONE, Год журнала: 2025, Номер 20(4), С. e0321011 - e0321011

Опубликована: Апрель 28, 2025

The rapid development of social media has significantly impacted sentiment analysis, essential for understanding public opinion and predicting trends. However, modality fusion in analysis can introduce a lot noise because the differences semantic representations among various modalities, ultimately impacting accuracy classification results. Thus, this paper presents Semantic Enhancement Cross-Modal Interaction Fusion (SECIF) model to address these issues. Firstly, BERT ResNet extract feature from text images. Secondly, GMHA mechanism is proposed aggregate important information mitigate influence noise. Then, an ICN module created capture complex contextual dependencies enhance capability representations. Finally, cross-modal interaction implemented. Text features are considered primary, image auxiliary, enabling profound integration textual visual features. model's performance optimized by combining cross-entropy KL divergence losses. experiments conducted using dataset collected events on Sina Weibo. results demonstrate that outperforms comparison models. SECIF improves 11.19%, 82.27%, 4.83% over average text-only, image-only, multimodal models, respectively. compared with ten baseline models publicly available datasets. experimental show 4.70% F1 score 6.56%. Through governments better understand emotions trends, facilitating more targeted effective management strategies.

Язык: Английский

---

A Comparative Study of Machine Learning and Deep Learning Approaches for Identifying Assamese Abusive Comments on Social Media

Procedia Computer Science, Год журнала: 2025, Номер 258, С. 981 - 992

Опубликована: Янв. 1, 2025

Язык: Английский

---

Harnessing Deep Learning for Assamese Sarcastic Comment Detection from Social Media Text

Procedia Computer Science, Год журнала: 2025, Номер 258, С. 3115 - 3125

Опубликована: Янв. 1, 2025

Язык: Английский

---

Leveraging Generative AI Synthetic and Social Media Data for Content Generalizability to Overcome Data Constraints in Vision Deep Learning

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Ноя. 28, 2024

Generalizing deep learning models across diverse content types is a persistent challenge in domains like Facial Emotion Recognition (FER), where datasets often fail to reflect the wide range of emotional responses triggered by different stimuli. This study addresses issue generalizability comparing FER model performance between trained on video data collected controlled laboratory environment, extracted from social media platform (YouTube), and synthetic generated using Generative Adversarial Networks. The videos focus facial reactions advertisements, integration these sources seeks address underrepresented advertisement genres, reactions, individual diversity. Our leverage Convolutional Neural Networks Xception architecture, which fine-tuned category based sampling. ensures training validation represent categories, while testing includes novel evaluate rigorously. Precision-recall curves ROC-AUC metrics are used assess performance. Results indicate 7% improvement accuracy 12% increase precision-recall AUC when combining real-world data, demonstrating reduced overfitting enhanced generalizability. These findings highlight effectiveness integrating build systems that perform reliably more representative content.

Язык: Английский

---

Lightweight Attention Based Deep CNN Framework for Human Facial Emotion Detection from Video Sequences

SN Computer Science, Год журнала: 2024, Номер 6(1)

Опубликована: Дек. 20, 2024

Язык: Английский

---