<![CDATA[This research developed a web-based real-time facial analysis system to overcome the challenge of detection accuracy in dynamic video streaming data. Using the face-api.js library with the Tiny Face Detector algorithm and a 68-point landmark model, the system is capable of simultaneously detecting faces, classifying gender, and recognizing seven basic emotional expressions. The main innovation of this system lies in the automatic extraction of five Regions of Interest (ROI) eyes, eyebrows, nose, mouth, and jaw and the presentation of confidence score data in the form of a time series graph. All analysis results are stored in a structured JSON dataset format for further research needs. The implementation results show high performance with an average confidence value above 90% in frontal face conditions and optimal lighting. The system has been proven to maintain detection stability up to a 30 degree face tilt angle and process data without significant latency. Although low light intensity can reduce the confidence value by 15-20%, this architecture proves the effectiveness of complex facial analysis using minimal hardware resources.]]>
Copyrights © 2026
