<![CDATA[The increasing demand for security systems has encouraged the use of image processing technology to assist in identifying potential criminal activities. One important indicator in security surveillance is the presence of individuals carrying weapons. Therefore, this study aims to implement the YOLO11 deep learning algorithm to detect and classify individuals based on the presence of weapons in images. The dataset used consists of two classes, namely Person and Weapon, and four training scenarios were evaluated: the original dataset, the MixUp-augmented dataset, the CutMix-augmented dataset, and the combined MixUp-CutMix dataset. Model performance was evaluated using precision, recall, mAP50, and mAP50-95 metrics. The results show that the model trained on the original dataset achieved the highest recall of 0.84 and mAP50 of 0.92, while the model trained on the combined MixUp-CutMix dataset achieved the highest precision of 0.95 and mAP50-95 of 0.62. Among the augmentation methods applied, the combination of MixUp and CutMix produced the best performance, whereas MixUp yielded relatively lower performance than the other methods. The findings indicate that data augmentation can increase the diversity of training data; however, it does not always lead to better performance than the original dataset. Overall, the results demonstrate that YOLO11 has strong potential for application in computer vision-based surveillance systems to support the automatic detection of individuals carrying weapons.]]>
Copyrights © 2026
