Abstract
The research paper provides a novel application of YOLOv8 for the detection of protection emblems in the International Federation of Red Cross and Red Crescent Societies (IFRC), aimed at enhancing humanitarian safety in conflict zones. The authors created a dataset of 36 images from various sources, with 25 used as training data, 4 as test data, and 7 as validation data. Custom annotation was performed through Roboflow, and YOLOv8 was employed for detection. The findings revealed a maximum F1 score of 0.9 at a confidence threshold of 0.85. The confusion matrix indicated a detection rate of 0.91 on real positive cases, particularly in the successful recognition of the class of significance, which is IFRC_Symbols. This research implemented object detection in a sensitive humanitarian context, connecting artificial intelligence (AI) with humanitarian operations, thereby helping to reduce the risk of misidentification of symbols during relief operations. In the preliminary stage, the authors used a small dataset, creating opportunities for future researchers to expand the dataset and implement the model in real-time scenarios, including other environmental factors. The decision to enhance emblem detection facilities aligns with the mission of the IFRC to assist people in need safely and efficiently without jeopardizing the efforts of its diligent employees.