Diseño e implementación de un prototipo de detección del estado de somnolencia para conductores mediante el uso del Software Teachable Machine

Abstract

This graduation project arises in response to the problem of traffic accidents, aiming to provide a solution targeted at preventing situations caused by driver drowsiness. A prototype has been developed, which is installed at the front of the dashboard, utilizing a camera to recognize images and trigger an alarm upon detecting signs of drowsiness. The implementation was carried out using the TEACHABLE MACHINE website, specifically in the image projects section. Through a webcam, images of poses associated with drowsiness were captured. Subsequently, a model was trained using the "Export Model" option of TEACHABLE MACHINE, employing the "Tensorflow" configuration and selecting "Saved Model". The trained model was downloaded in .h5 format, which was then integrated into a Raspberry Pi. In the final phase of the project, the .h5 file was imported into the Raspberry Pi, and a port was selected to connect a speaker. This speaker emits a sound alarm upon recognizing characteristic poses of drowsiness, thereby alerting the driver about their state. In summary, this work focuses not only on technical development using tools such as TEACHABLE MACHINE and Raspberry Pi but also on the application of this knowledge to enhance road safety, especially in situations involving driver fatigue.