Akademik Veri Yönetim Sistemi
9th International Artificial Intelligence and Data Processing Symposium, IDAP 2025, Malatya, Türkiye, 6 - 07 Eylül 2025, (Tam Metin Bildiri)
Peak frame selection is a critical task in video analysis, particularly in applications such as emotion recognition, where identifying the most informative frames from continuous video sequences can significantly enhance both computational efficiency and classification accuracy. In this study, a novel approach to peak frame selection is proposed, setting it apart from existing methods in the literature through its unique integration of YOLOv8-based face detection with Canny edge detection for expression change localization. Initially, facial regions were extracted from video frames using YOLOv8 to ensure accurate and reliable face detection. Subsequently, the Canny edge detection algorithm was used to identify frames that exhibit the most pronounced changes in facial expression. To evaluate the accuracy and consistency of the selected peak frames, classification analyzes were performed using pretrained GoogLeNet and MobileNet architectures. The integration of YOLOv8 and Canny edge detection offers a robust and effective solution for selecting peak frames, thereby contributing to improved emotion recognition performance from facial video data.