Driver inattention is a leading cause of road accidents, with distracted drivers engaging in anomalous behaviors at a higher risk of causing potentially deadly incidents. This highlights the critical need for effective Driver Monitoring Systems (DMS) to improve road safety. By continuously monitoring and analyzing a driver’s actions, DMS can help identify and address attention lapses, ensuring safer driving conditions and ultimately saving lives. This study evaluates the top-performing deep learning models identified in the literature, specifically CNN, MobileNet, ResNet, and YOLO, in detecting driver inattention within a unimodal framework using the DMD-TFIW0 dataset. Evaluation based on classification metrics and computational efficiency reveals that ResNet provides the optimal balance, achieving 95.33% accuracy in multiclass classification (six driver states) and 96.46% in binary classification (safe vs. anomalous driving), with processing speeds of 0.11 ms/frame and 0.10 ms/frame, respectively. These results demonstrate ResNet’s potential for real-time, in-vehicle driver inattention detection and its contribution to safer driving.