Alertness of the driver behind the wheel in the vehicle is of paramount importance. Driver drowsiness can induce accidents and endanger their life along with other persons as well. Although there were works for developing devices from health sensors to gauge the driver alertness, they could sometimes be physically intrusive and may cause physical discomfort to the driver. Accurate detection of tired drivers based on physiological data may be slow to raise an alert and sometimes could be unreliable. This paper takes advantage of the behavioral changes in the driver for early detection of drowsiness among drivers. This work trains a convolutional neural network that observes various behavioral changes in the driver to assess the alertness of the driver. If the driver is drowsy, then an alarm mechanism alerts driver to take safety measures. Among the three different deep learning methods used, YOLO v9 has provided the best accuracy for the given dataset.