Image emotion classification leveraging a fusion technique combining LBP and YOLO

Understanding human emotions is a topic of interest for researchers in various fields because emotional factors often affect daily life or may reduce work efficiency due to a lack of emotional readiness. The study of emotion therapy through images can be applied to solve mental health problems, starting with the study of Visual Emotion Analysis (VEA) and the classification of emotions from images to identify the emotional states or feelings conveyed through image content. It is a technology with a wide range of applications, from improving the user experience in applications to examining emotional responses in various contexts. To accurately and effectively classify emotions in images, especially since a single image can contain multiple hidden emotions, researchers have proposed guidelines to improve process as a novel approach to classifying emotions in images by combining Local Binary Pattern (LBP) and YOLOv8 techniques. These methods extract features from both objects and textures in images. A new model has been designed by modifying the last layer of ResNet-50 and adding a new layer to integrate features from the enhanced LBP histogram to handle the combined features from LBP and YOLOv8. The results show that the model can classify emotions in images accurately and is robust to changes in lighting, perspective, and cropping, achieving up to 98% accuracy. The model is suitable for real-world applications and further development in the field of emotion classification from images.