Integração de Inteligência Artificial e Soluções IoT para Automatização dos Serviços de Empréstimo de Bicicletas Elétricas

Abstract

This work proposes the creation of an automated platform for electric bicycle rentals, integrating Artificial Intelligence (AI) and Internet of Things (IoT) technologies to simplify use and enhance the security of this service. By using an AI-based facial recognition system connected to a WhatsApp bot, users can rent bicycles without the need for additional apps. The literature review explores the main AI architectures evaluated for the project, including Fully Convolutional Networks (FCN), Mask R-CNN, U-Net, and DeepLabV3+ for segmentation, and ResNet50, EfficientNetB3, Visual Geometry Group (VGG), and MobileNet for image classification, highlighting their applications and advantages. The methodology covers the development of the system in Python, using frameworks like FastAPI and AI libraries (TensorFlow, Scikit-learn, OpenCV, and Keras), employing a robust model trained for face segmentation and training customized classification models for each user. The results show that the DeepLabV3+ segmentation architecture, with 94.64% accuracy and 12.19% loss during validation, and an overall accuracy of 93%, and the ResNet50 classification model, which achieved 100% accuracy and 0.66% loss during validation, maintaining an accuracy of 98% to 100% among models trained for different faces, offered the best performance in terms of accuracy and efficiency for facial recognition. The main system was able to perform processes such as training customized user models in an average of 17 minutes during registration, validation processes in an average of 10 seconds, and secondary processes in 2 seconds. The integration between the WhatsApp bot and the main system's API proved effective in managing user data flow, meeting the security and practicality requirements of the service. It is concluded that the developed system contributes to the secure and simplified automation of urban mobility services, with potential for future expansions into other areas of automation.