MILOUD ABID AbirBAHAOUS Chaimaa Fatima ZahraBENTAIEB Samia2026-09-082026-09-082026https://dspace.univ-temouchent.edu.dz/handle/123456789/7474This work aims to develop an intelligent system for detecting driver drowsiness while driving, with the goal of enhancing road safety through direct monitoring. This project oers an embedded solution based on computer vision and deep learning techniques. Our methodology is based on a comparative evaluation between two methods: the rst model relied on the MobileNet architecture, which showed a quick response during the training phase but did not achieve the required eectiveness during actual operation. Consequently, the second model was developed using MediaPipe technology, which proved highly accurate in studying facial landmarks during practical operation, making it the optimal choice for this system. The system's eectiveness was veried through practical tests in real driving conditions inside a vehicle, coupled with an audible alert system for immediate warning. The results highlight that relying on MediaPipe technology represents the most ecient and reliable solution, eectively contributing to the development of driver monitoring techniques and the reduction of trac accidents caused by fatigue.frDrowsiness detectioncomputer visionMobileNetMediaPipeembedded systemsdriver safety.Thesis