One of the main objectives of Artificial Intelligence and Computer Vision is to totally avoid road accidents caused by tired drivers by developing drowsiness detection systems. By using Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM), the system monitors and detects drowsiness. Convolutional Neural Networks (CNNs), is a kind of deep learning technique which considers driver’s facial features and based on these features, our system is able to detect fatigue in real time. The technology captures photos or video frames of a driver’s face and analyzes key facial expressions, head positions, as well as eye shut to determine driver’s overtiredness. We have used a variety of dataset which we have collected in real time. This contains both active and drowsy facial expressions. A CNN model that has been trained using this dataset achieved high accuracy. It uses IR camera with addition of Haar Cascade and Dlib integration to detect landmarks under varying light conditions. The model which we have proposed in this study recognizes the driver’s state, analyzes visual cues, and extracts spatial characteristics by adopting sophisticated deep learning architectures. By combining optimized tactics, real-time monitoring, and image processing approaches, the proposed system achieves good reliability. This study has important implications for road safety, fleet management, and smart vehicle systems. Our proposed system will start alarming immediately when the sleepiness is detected on driver’s face. In this way, the model helps the passengers as well as drivers onboard by avoiding road accidents which in turn increase road safety. The proposed work has a greater potential to further extend to multi-modal systems with the addition of physiological signals and reinforcement learning which could improve drowsiness detection more precisely and make transportation more reliable.