Real-time face emotion recognition and visualization using grad-CAM

Abstract

One of the most indicative ways of communication is facial expressions. The Face attributes are the contended mode to specify human sensitivity. Hence facial emotion recognition is necessary for human-machine interaction systems. The AI nowadays can also understand emotions verifying facial movement and intimation like a human brain does. But tracing the mechanism of AI is challenging as most of the AI methods are referred to as "Black box". To perceive the insights of AI algorithms the term Explainable AI has been brought to light. Explainable AI is a need to implement and build proper, fair, and responsible models that can even be flexible to use on a large production basis. In this paper, Cascade Classifier for emotion recognition and Grad-CAM for visualization of model detection has been employed. The region of interest of the face is located to extract features which are categorized into 7 classes. The results obtained are appreciable and can be applied in works relating to human expression detection.