In hand-based biometric identification and authentication systems, the localization of the knuckles is a crucial task. The accurate localization of knuckle regions in images has shown to have great potential in recent years thanks to deep learning techniques. The YOLOv5 algorithm, a cutting-edge object detection algorithm, has attracted a lot of attention in this regard because of its high accuracy and quick processing time. In this study, we use YOLOv5 to propose a novel method for knuckle localization. Training and testing are the two main phases of the suggested strategy. We trained the YOLOv5 algorithm using a sizable dataset of hand images during the training phase. To locate the knuckle regions in the images, the dataset was manually annotated. We then improved the pre-trained YOLOv5 model’s localization of knuckles accuracy using our dataset. During the testing phase, we assessed the effectiveness of our strategy using a different test dataset. Performance was evaluated using the mean intersection of union (mIoU) and mean average precision (mAP) measures. Our recommended approach for knuckle localization using YOLOv5 demonstrates the strength and capability of deep learning techniques in solving knuckle localization issues. We use this as a future work, a comparative examination of the holistic (all twelve knuckles of 4 fingers) knuckle localization problem will be discussed.