The increasing deployment of smart devices in residential environments opens new opportunities for intelligent energy management. However, existing platforms often fall short in providing intuitive interfaces, zone-level control, and advanced predictive analytics accessible to non-expert users. This paper presents the design of a modular smart home energy management system that integrates real-time monitoring, consumption forecasting, and intelligent assistance via Large Language Models (LLMs). The system features an interactive floor plan interface, multi-user support, threshold-based alerting, and detailed historical analytics. Additionally, it introduces LLM-powered agents that guide users in configuring smart devices and adopting more efficient consumption behaviors. This architecture emphasizes accessibility, adaptability, and extensibility, aiming to empower users with actionable insights and seamless device management. The proposed solution addresses current gaps in existing platforms and lays the groundwork for intelligent, personalized, and proactive home energy systems.

The development of a crowd digital twin offers significant potential for enhancing public safety, urban planning, and event management. A key challenge in creating such a digital twin lies in the efficient and accurate acquisition of crowd-related data, particularly through object detection models deployed on resource-constrained devices. Through a series of experiments, we compare TinyML and Edge approaches in terms of detection accuracy, inferencing time, and resource utilization. Our findings highlight the trade-offs inherent in selecting detection models for crowd digital twin applications, underscoring the importance of aligning model choice with specific deployment needs.

Vision equipment plays a crucial role in wildlife conservation by enabling the detection and collection of wild animal images, thereby providing an efficient way for preserving biodiversity observation. However, traditional manual detection methods are inefficient and costly. While cloud server-based methods offer an alternative, they introduce challenges such as transmission delays and data security concerns. To address these limitations, we propose an edge computing-based AI vision terminal for autonomous wildlife monitoring. Evaluations using the NCNN framework and varying input resolutions (640,320, 160 pixels) revealed that YOLOv8n models resulted in significantly faster inference times (up to 7.8-14.5x speedup at 160 pixels compared to 640 pixels). We implemented and evaluated quantized YOLOv8n and YOLOv8s models using NCNN on a Raspberry Pi, achieving significant inference speedups (18.840.1% reduction in inference time) compared to non-quantized models across various image sizes. YOLOv8s-int8 offered a better speed-precision trade-off ($24 \%$ faster for $6.8 \%$ lower precision) than YOLOv8n-int8 ($\mathbf{1 2 . 8 \%}$ faster for $\mathbf{1 1 . 6 \%}$ lower precision). This approach enables real-time animal detection with approximately 3 W power consumption, demonstrating the feasibility of deploying intelligent wildlife monitoring systems in remote, resource-constrained environments. Furthermore, the edge device exhibits robust detection performance for complex backgrounds and small targets.

Real-world handwritten documents often contain noise and complex elements, such as notes with colored markings, naturally degraded handwriting, and diverse paper backgrounds. Based on the strong demand for techniques that convert text images into editable digital formats, this study focuses on recognizing line-level handwritten Chinese text in complex backgrounds to improve recognition accuracy. Through a comparative analysis of five experimental settings, including no preprocessing, different preprocessing techniques, and advanced enhancement methods leveraging the self-attention mechanism from the transformer network, our reimagined CRNN model achieves the highest accuracy. These results confirm the effectiveness of the selfattention mechanism in boosting recognition performance under challenging conditions, offering valuable insights for future advancements in handwritten text recognition technologies.

Modality differences and intramodality variations make the visible-infrared person reidentification (VI-ReID) task highly challenging. Most existing methods focus on building network frameworks based on convolutional neural networks (CNN) or pure vision transformers (ViT) to extract discriminative features and address these challenges. However, these methods neglect several key issues: deeply fusing local features with global spatial information enhances comprehensive discriminative representation, patch tokens contain rich semantic information, and different feature extraction stages within the network emphasize various semantic elements. To address these issues, we propose a novel hybrid ResNet-transformer hierarchical aggregation architecture named HAResformer. HAResformer comprises three key components: 1) hierarchical feature extraction (HFE) framework; 2) deeply supervised aggregation (DSA); and 3) hierarchical global aggregate encoder (HGAE). Specifically, HFE introduces a lightweight cross-encoder feature fusion module (CFFM) to deeply integrate the local features and global spatial information of a person extracted by the ResNet encoder (RE) and transformer encoder (TE). Subsequently, the fused features are fed as global priors into the next-stage TE for deep interaction, aiming to extract specific local features and global contextual clues. Additionally, DSA and HGAE provide auxiliary supervision and aggregation on multiscale features to enhance multigranularity feature representation. HAResformer effectively alleviates modality differences and reduces intramodality variations. Extensive experiments on three benchmarks demonstrate the effectiveness and generalization of our architecture and outperform most state-of-the-art methods. HAResformer has the potential to become a new VI-ReID baseline, promoting high-quality research in the future.

Multimodal interfaces go beyond the traditional interaction through keyboard and mouse by incorporating multiple modes of interaction, such as touch, voice, gesture, and even gaze, to create more intuitive and immersive user experiences. This paper investigates how TinyML can be employed for multimodal interfaces in the context of games. An endless game in which the character has to avoid obstacles and fight monsters to advance has been developed. An Arduino Nano 33 BLE Sense is then used as the input device for the game by recognizing the hand gestures of the players.

A key challenge of visible-infrared person re-identification (VI-ReID) comes from the modality difference between visible and infrared images, which further causes large intra-person and small inter-person distances. Most existing methods design feature extractors and loss functions to bridge the modality gap. However, the unpaired-images constrain the VI-ReID model's ability to learn instance-level alignment features. Different from these methods, in this paper, we propose a pose attention-guided paired-images generation network (PAPG) from the standpoint of data augmentation. PAPG can generate cross-modality paired-images with shape and appearance consistency with the real image to perform instance-level feature alignment by minimizing the distances of every pair of images. Furthermore, our method alleviates data insufficient and reduces the risk of VI-ReID model overfitting. Comprehensive experiments conducted on two publicly available datasets validate the effectiveness and generalizability of PAPG. Especially, on the SYSU-MM01 dataset, our method accomplishes 7.76% and 5.87% gains in Rank-1 and mAP. The code is available athttps://github.com/qyhsxdx/PAPG.

Monitoring electric vehicles' battery status and forecasting their state of health is still an open challenge. To determine how and why a battery degrades over time, we have extensively monitored a Nissan Leaf's battery pack for more than one year. Collecting more than 4.5 million samples via a custom monitoring connected device to investigate how different driving behaviors affect battery aging. In addition, the best driving behaviors based on the battery's optimal temperature are revealed, including speed, acceleration and brake pedal pressure, and horsepower.

The Internet Computer Protocol is a new generation blockchain that aims to provide better security and scalability than the traditional blockchain solutions. In this paper, this innovative distributed computing architecture is introduced, modeled and then simulated by means of an agent-based simulation. The result is a digital twin of the current Internet Computer, to be exploited to drive future design and development optimizations, investigate its performance, and evaluate the resilience of this distributed system to some security attacks. Preliminary performance measurements on the digital twin and simulation scalability results are collected and discussed. The study also confirms that agent-based simulation is a prominent simulation strategy to develop digital twins of complex distributed systems.

COVID-19 pandemic has led to serious economic and life losses. Face Masks serve as first infection barrier when used in public spaces. In this paper, we propose a new near-realtime method to automatically recognize face mask wearing that combines human posture recognition with convolutional neural network (CNN). We use the power of human posture recognition to perform background filtering and spatial reduction in the original images. The outcome is then used by a trained CNN model to identify if the subject is wearing a mask. We exploit Openpose to identify the skeleton of human body and locate the facial region thus spatially reducing the area to be processed by the CNN framework. We then adopt supervised learning approach to detect if a face mask is present. The CNN is trained using images, cropped to the supposed face mask covered region. This approach led to a substantial reduction in neural network complexity yet improving the recognition accuracy. The system has been evaluated in a multitude of scenarios using images taken in public places at different time of day and with different angles. Overall, our system achieves a recognition accuracy of 95.8% and 94.6% in daytime and nighttime respectively.

Humans generally spend more than 90% of their time in indoor environments. Such value can reach 100% due to the restrictions and limitations we are experiencing because of the current COVID-19 pandemic. Indeed, monitoring the indoor air quality (IAQ) becomes strategic to prevent and limit risks and adverse effects on building occupants’ health, comfort, and well-being. To reduce one of the variables impacting the IAQ, i.e. CO2, indoor plants can be exploited. In this paper, we present the low-cost prototype of a system we designed with the intent to foster user’s awareness about IAQ exploiting an IoT-enabled home gardening system, able to sense information about the ambient conditions and plant health.