Facial recognition is a computer application that can identify, track, or verify human faces based on images or videos taken with a digital camera. Although significant advances have been made in facial recognition and security, identity authentication, and presence detection, there are still challenges that prevent human-level accuracy from being surpassed. These problems are variations on the appearance of the human face, such as various lighting conditions, face image noise, scale, pose, and others. This research paper presents a new approach by using CNN and Haar Cascade classifiers to solve some of these issues and improve the accuracy and detection rate of faces. While most tutorials on the topic aim to improve the accuracy of facial detection, this presented methodology allows facial recognition to detect and recognize frontal, profile photos as well as partially occluded faces. This unique feature of the proposed model opens up various applications in real-world scenarios where faces may be partially obscured, such as in crowded classrooms, busy office environments, or even in public spaces where people may attempt to break traffic signals. Experimental results show that the proposed method is comparable with the latest models and can be implemented in a real-world environment as an automated system.