arXiv:2511.10940v1 Announce Type: new 
Abstract: Facial Expression Recognition remains a challenging task, especially in unconstrained, real-world environments. This study investigates the performance of two lightweight models, YOLOv11n and YOLOv12n, which are the nano variants of the latest official YOLO series, within a unified detection and classification framework for FER. Two benchmark classification datasets, FER2013 and KDEF, are converted into object detection format and model performance is evaluated using mAP 0.5, precision, recall, and confusion matrices. Results show that YOLOv12n achieves the highest overall performance on the clean KDEF dataset with a mAP 0.5 of 95.6, and also outperforms YOLOv11n on the FER2013 dataset in terms of mAP 63.8, reflecting stronger sensitivity to varied expressions. In contrast, YOLOv11n demonstrates higher precision 65.2 on FER2013, indicating fewer false positives and better reliability in noisy, real-world conditions. On FER2013, both models show more confusion between visually similar expressions, while clearer class separation is observed on the cleaner KDEF dataset. These findings underscore the trade-off between sensitivity and precision, illustrating how lightweight YOLO models can effectively balance performance and efficiency. The results demonstrate adaptability across both controlled and real-world conditions, establishing these models as strong candidates for real-time, resource-constrained emotion-aware AI applications.

تدرس دراسة مقارنة حول التعرف على تعبيرات الوجه أداء نموذجين خفيفين، YOLOv11n وYOLOv12n. تم اختبار النماذج باستخدام مجموعات البيانات FER2013 وKDEF، التي تم تحويلها إلى تنسيقات كشف الكائنات. تشير النتائج إلى أن YOLOv12n حقق متوسط دقة (mAP) بنسبة 95.6% على مجموعة بيانات KDEF، متفوقًا على YOLOv11n الذي سجل 63.8% على FER2013. ومع ذلك، أظهر YOLOv11n دقة أعلى بنسبة 65.2%، مما يدل على موثوقيته في البيئات المليئة بالضوضاء.

Un estudio comparativo sobre el reconocimiento de expresiones faciales (FER) evalúa el rendimiento de dos modelos ligeros, YOLOv11n y YOLOv12n. Los modelos fueron probados utilizando los conjuntos de datos FER2013 y KDEF, convertidos a formatos de detección de objetos. Los resultados indican que YOLOv12n logró una media de precisión (mAP) del 95.6% en el conjunto de datos KDEF, superando a YOLOv11n, que obtuvo un 63.8% en FER2013. Sin embargo, YOLOv11n demostró una mayor precisión del 65.2%, lo que indica su fiabilidad en entornos ruidosos.

Une étude comparative sur la reconnaissance des expressions faciales (FER) évalue la performance de deux modèles légers, YOLOv11n et YOLOv12n. Les modèles ont été testés à l'aide des ensembles de données FER2013 et KDEF, convertis en formats de détection d'objets. Les résultats indiquent que YOLOv12n a atteint une moyenne de précision (mAP) de 95,6 % sur l'ensemble de données KDEF, surpassant YOLOv11n, qui a obtenu 63,8 % sur FER2013. Cependant, YOLOv11n a démontré une précision plus élevée de 65,2 %, indiquant sa fiabilité dans des environnements bruyants.

A comparative study on Facial Expression Recognition (FER) evaluates the performance of two lightweight models, YOLOv11n and YOLOv12n. The models were tested using the FER2013 and KDEF datasets, converted into object detection formats. Results indicate that YOLOv12n achieved a mean Average Precision (mAP) of 95.6% on the KDEF dataset, outperforming YOLOv11n, which scored 63.8% on FER2013. YOLOv11n, however, demonstrated higher precision at 65.2%, indicating its reliability in noisy environments.

Facial Expression Recognition with YOLOv11 and YOLOv12: A Comparative Study

Was this article worth reading? Share it

Ready to build your own newsroom?