Processant ...

This project evaluates the technical viability of integrating artificial intelligence, specifically computer vision, within the paradigm of edge computing for the Internet of Things (IoT) applications. The inquiry is motivated by the current limitations of cloud-centric AI models, such as high latency, bandwidth consumption, and energy demands, which render them unsuitable for many telecommunications engineering scenarios that require efficient, low-cost, and autonomous operation. The methodological framework is based on the theoretical principles of image processing and computer vision, notably drawing from the teachings of Dr. Domènec Puig on biomedical image analysis. A practical case study was developed using an NVIDIA Jetson Nanoplatform, implementing a real-time facial recognition system for access control. The system leverages OpenCV for face detection and the Dlib library for feature extraction and encoding comparison, operating under an edge-computing architecture with optional cloud-based notification via Telegram. The results demonstrate the functional feasibility of deploying computer vision on resource-constrained hardware, although challenges such as susceptibility to spoofing and recognition variability under different angles and lighting conditions were identified. The study concludes that while current AI models can be adapted to edge-IoT contexts, significant improvements in robustness and efficiency are still needed for widespread commercial adoption. This work contributes to the ongoing effort to decouple AI performance from cloud dependency and supports the development of more sustainable and decentralized intelligent systems.