LAUSR

Khorat snail—eating turtle ( Malayemys khoratensis ) is widely recognized as a valuable bioindicator for environmental monitoring due to its long life span and high accumulative capacity. Until now, specialized techniques such as carapace marking and microchip implantation are required for thelong-term tracking of these freshwater turtles. However, these techniques are costly and frequently cause physical harm to the turtles, disrupting their natural behaviors and feeding patterns. To overcome these circumstances, this study explores the precision of YOLOv8 for the individual identification of Khorat snail-eating turtles through the analysis and finding of characteristic patterns, including nasal stripes, position and shape of the infraorbital stripe, and plastron stripe patterns and enhancing by directional weight parameter to detect individual and recognition accurately. A directional weight parameter was introduced into the YOLOv8 based on convolutional neural network (CNN) framework to improve recognition performance. This parameter emphasizes the spatial orientation of key morphological features such as plastron stripes and facial markings, thereby guiding the model to focus on biologically stable and directionally significant traits during individual identification. Among the proposed models, the utilizing a novel approach based on plastron stripe pattern demonstrated the highest accuracy, achieving up with a precision of 0.96 ± 0.01, recall of 0.97 ± 0.01, accuracy of 0.97 ± 0.01, and a mAP@50–95 of 0.91 ± 0.01 precision in individual identification when trained on 5-fold cross-validation of data collect from 30 Khorat snail-eating turtles. The findings highlight that biometric identification based on morphological traits can be used for individual identification of Khorat snail-eating turtles, and it should be considered a choice for noninvasive long-term tracking of these turtles. Utilizing YOLOv8 techniques to identify distinctive patterns for individual recognition in the Khorat snail-eating turtle ( Malayemys khoratensis ) presents significant potential for enhancing the accuracy and efficiency of ecological monitoring efforts. This approach could substantially improve individual identification accuracy, thereby contributing to more effective monitoring and research in ecological studies.