Pedro Jorge Sequeira Cardoso

Pose-based video anomaly detectors can only score what the pose estimator detects: frames with no detected person are scored as maximally normal, and anomalies caused by non-human objects produce no signal at all. We show that a frozen vision–language model (VLM), queried only where the pose model is weak, mitigates this blind spot without retraining either component. Keeping the Spatio-Temporal Graph Normalising Flow (STG-NF) frozen, a label-free router sends the 3.7% of frames on which it is unreliable to the VLM, and a category-aware, positive-only fusion lets the VLM raise the anomaly score but never lower it. On ShanghaiTech this raises the cross-fitted global ROC-AUC from 0.8594 to 0.8857 (Δ = +0.0263; 95% CI [+0.0085, +0.0483]), improving the human-related subset as well, for about five euros of inference. A frozen object detector under an identical pipeline recovers less than half the gain.

Tourism, cultural events, and trade exhibitions increasingly need real-time, privacy-risk-reducing methods for quantifying visitor behavior and satisfaction at physical stands. This article addresses a fragmented evidence base by (i) presenting a focused, design-oriented literature synthesis of multimodal sensing for stand-like environments and (ii) proposing a unified, deployment-oriented reference architecture. It analyses a 60-paper design corpus (2022–2026), together with selected benchmark/context references across RGB-D, thermal cameras, LiDAR, mmWave radar, microphone arrays, and multimodal fusion, and maps the analytic corpus to event-relevant key performance indicators (KPIs). The synthesis identifies gaps in the corpus: scarce stand-specific, multimodal KPI-labeled event datasets, conceptual inconsistencies in satisfaction inference, and limited direct evidence for commercially important KPIs such as feedback, conversion, and retention. In response, the article introduces the Stand Multimodal Behavior and Satisfaction Analysis (SMBSA) architecture, a five-layer edge-AI conceptual reference architecture with hierarchical fusion and candidate analytics heads for future validation of behavior and satisfaction KPIs under GDPR-aware, privacy-by-design requirements.

The transition towards sustainable educational campuses requires robust energy management strategies that integrate operational oversight with advanced analytics. This paper presents a campus-scale electricity monitoring system at the University of Algarve, designed to support the institution’s sustainability goals through continuous monitoring, data reliability, portability, and scalability to handle concurrent high-frequency campus-wide telemetry. The system consolidates heterogeneous meters into a unified platform, enabling precise tracking of energy consumption and photovoltaic generation. Beyond operational efficiency, the platform incorporates a data-driven analytical layer featuring short-term forecasting using Prophet, chosen for its computational scalability, and a hybrid anomaly detection scheme combining forecast residuals with Isolation Forest. These capabilities facilitate the early identification of waste and abnormal consumption patterns, directly contributing to energy conservation and carbon footprint reduction. Validated across multiple buildings, the system demonstrates both portability to different energy profiles and high data continuity, reducing the cognitive load on facility managers. By providing a reproducible blueprint for intelligent energy monitoring, this work supports institutions in their pursuit of energy efficiency and sustainable development, aligning operational practices with broader environmental objectives.

The transition towards sustainable educational campuses requires robust energy management strategies that integrate operational oversight with advanced analytics. This paper presents a campus-scale electricity monitoring system at the University of Algarve, designed to support the institution's sustainability goals through continuous monitoring, data reliability, and scalability. The system consolidates heterogeneous meters into a unified platform, enabling precise tracking of energy consumption and photovoltaic generation. Beyond operational efficiency, the platform incorporates a data-driven analytical layer featuring short-term forecasting using Prophet and a hybrid anomaly detection scheme combining forecast residuals with Isolation Forest. These capabilities facilitate the early identification of waste and abnormal consumption patterns, directly contributing to energy conservation and carbon footprint reduction. Validated across multiple buildings, the system demonstrates high data continuity and effective anomaly detection, reducing the cognitive load on facility managers. By providing a reproducible blueprint for intelligent energy monitoring, this work supports institutions in their pursuit of energy efficiency and sustainable development, aligning operational practices with broader environmental objectives.

The accurate measurement of audience engagement in real-world live events remains a significant challenge, with the majority of existing research confined to controlled environments like classrooms. This paper presents a comprehensive survey of Computer Vision AI-driven methods for real-time audience engagement monitoring and proposes a novel, holistic architecture to address this gap, with this architecture being the main contribution of the paper. The paper identifies and defines five core constructs essential for a robust analysis: Attention, Emotion and Sentiment, Body Language, Scene Dynamics, and Behaviours. Through a selective review of state-of-the-art techniques for each construct, the necessity of a multimodal approach that surpasses the limitations of isolated indicators is highlighted. The work synthesises a fragmented field into a unified taxonomy and introduces a modular architecture that integrates these constructs with practical, business-oriented metrics such as Commitment, Conversion, and Retention. Finally, by integrating cognitive, affective, and behavioural signals, this work provides a roadmap for developing operational systems that can transform live event experience and management through data-driven, real-time analytics.