Facial emotion recognition constitutes a central challenge in computer vision, where the objective is to infer affective states from subtle and often ambiguous facial patterns under real world conditions. In this work, we develop a deep learning–based framework capable of performing real time emotion classification while preserving high predictive accuracy and computational efficiency.Building on a detailed analysis of state of the art facial emotion recognition architectures, we implement a fine tuned ResNet18 model optimized for low latency inference and robust feature extraction.This configuration achieves an effective compromise between representational capacity and model complexity, enabling deployment on resource constrained platforms. The resulting system demonstrates strong applicability across diverse vision centric domains including embedded perception modules, clinical monitoring pipelines, and adaptive learning environments highlighting the operational relevance and scalability of modern emotion recognition approaches.
- Efficient Real-Time Facial Emotion Recognition via Optimized ResNet18 Feature Extraction
- Hamza Graïn - IMT Mines AlèsYoussef Hayani - IMT Mines AlèsYanis Blot-El Mazouzi - IMT Mines AlèsAmadou Diallo - IMT Mines AlèsHamza Bayd - Université de Montpellier, EuroMov DHM - EuroMov Digital Health in MotionImène Sekkiou - Caplogy InnovationBaptiste Magnier - Université de Montpellier, EuroMov DHM - EuroMov Digital Health in Motion
- ICFSP 2026 - 11th International Conference on Frontiers of Signal Processing (Toulouse, France, 10/06/2026–12/06/2026)
- 99260656709311
- EuroMov DHM - EuroMov Digital Health in Motion
- English
- Conference proceeding
- hal-05678688