Title: Real-Time Facial Emotion Detection via Supervised Contrastive Learning in a Multiscale Network

Abstract: Identifying emotions in real-time from facial expressions remains a complex endeavor, especially within video contexts where emotional states shift dynamically over time. This challenge is compounded by the fact that the facial manifestations of specific emotions differ markedly from person to person. Furthermore, emotional expression is not a series of discrete jumps but a continuous evolution, making it difficult to capture using traditional computational methods. A system capable of accurately tracking these subtle variations in facial dynamics could greatly enhance the assessment of an individual’s emotional condition. Such technology holds particular promise for psychological counseling, offering professionals deeper insights into their clients' emotional landscapes. This paper introduces a deep learning framework designed to identify emotional transitions in live video footage by explicitly modeling the continuity of facial expression changes. The proposed approach was evaluated on a standard dataset, yielding highly satisfactory performance results.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC