This paper presents a comprehensive approach to building energy efficiency analysis by integrating outdoor drone thermal imagery with real-time indoor heat envelope monitoring using deep learning techniques. The study builds upon our previous work that employed a YOLOv4 model to identify heat envelopes from drone-captured thermal images. To enhance the scope and robustness of the analysis, we introduce a complementary indoor monitoring system that utilizes a thermal camera for capturing real-time images of interior building spaces. Due to the scarcity of labeled indoor thermal data, we employ web scraping techniques to collect a diverse dataset for model training. The trained model is then evaluated on real-time indoor thermal images, demonstrating its effectiveness in identifying heat envelopes and potential energy inefficiencies. The combination of outdoor and indoor monitoring provides a holistic view of building energy performance, enabling targeted interventions for improved efficiency. This research advances the field of energy efficiency analysis by leveraging state-of-the-art deep learning algorithms and diverse data sources, offering a scalable and data-driven approach to building energy management. The proposed methodology has significant implications for reducing energy consumption, costs, and environmental impact in both residential and commercial buildings.
The Electronic Multi-User Randomized Circuit Training (EMURCT, pronounced “emmersed”) system has been developed to utilize pervasive computing and communication technology to address the lack of motivation that individuals have for exercising regularly. EMURCT is capable of producing a totally different workout sessions with every use, for up to 7 trainees simultaneously using a common workout circuit, in an effort to reduce boredom. EMURCT is composed of three different components, a client application, an administrator application and a web service. This project also uses Wi-Fi signal strength to estimate the activity level of individuals using the system. Based on the activity level being read from each smart device, the event scheduler has the option of releasing any trainee from his/her assigned station if it feels that he/she is not working out, and someone else’s smart device is requesting that station. Initial experiments indicate that the best signal strength reading comes from preset, dedicated access points.