Real-time face and smile detection has become an essential component of interactive multimedia systems that require natural and responsive human–computer interaction. However, achieving an optimal balance between detection accuracy, processing speed, and robustness under varying lighting, pose, and environmental conditions remains a significant challenge, particularly for systems intended to run on standard hardware with limited computational resources. This study aims to develop a real-time face and smile detection system using the Haar Cascade method based on OpenCV and implemented through a web-based interface built with Streamlit. The proposed system applies a cascaded detection approach, where face detection is performed on the entire video frame, followed by smile detection restricted to the detected facial Region of Interest (ROI). This strategy improves computational efficiency while reducing false positive detections. Performance evaluation was conducted using the Frames Per Second (FPS) metric, calculated dynamically from the processing time between consecutive frames. Experimental results indicate that the system operates effectively in real-time conditions, maintaining performance at approximately 5 FPS under moderate computational load, while still providing accurate visual annotations in the form of bounding boxes and labels.The system demonstrates high efficiency, ease of implementation, and acceptable accuracy in controlled environments. Nevertheless, performance degradation is observed under low lighting, non-frontal poses, and partial occlusions. These findings highlight the need for future improvements through the integration of deep learning-based approaches to enhance adaptability and detection robustness.
You may also start an advanced similarity search for this article.