Identification of fingerprint sweat hole images in forensic science based on YOLOV10

Fingerprint sweat holes as an important physical evidence in criminal technology has a high evidence value, but the traditional fingerprint sweat holes comparison mainly relies on manual, the image feature extraction is difficult, the recognition speed is slow and the accuracy is low and other problems. In this paper, we use the latest YOLOv10 image recognition model to recognize and classify images, and the fingerprint sweat holes construct the fingerprint sweat holes dataset, and evaluate the speed and accuracy performance of the different models, to explore and study applicable to the deep learning model recognition fingerprint sweat holes. The detection effect is better than traditional models such as SSD, Faster R-CNN, YOLOv5, YOLOv8 etcof YOLOv10 model on fingerprint sweat holes image dataset, in which the accuracy, recall, mAP50, mAP50-95 of Yolov10-X model reaches 99.4%, 99.1%, 99.5%, 96.7%. , but the training time and detection speed are the fastest for YOLOv10-S, which is more suitable for mobile smart devices.