Preliminary Development of Global-Local Balanced Vision Transformer Deep Learning with DNA Barcoding for Automated Identification and Validation of Forensic Sarcosaphagous Flies.

Morphological classification is the gold standard for identifying necrophilous flies, but its complexity and the scarcity of experts make accurate classification challenging. The development of artificial intelligence for autonomous recognition holds promise as a new approach to improve the efficiency and accuracy of fly morphology identification. In our previous study, we developed a GLB-ViT (Global-Local Balanced Vision Transformer)-based deep learning model for fly species identification, which demonstrated improved identification capabilities. To expand the model's application scope to meet the practical needs of forensic science, we extended the model based on the forensic science practice scenarios, increased the database of identifiable sarcosaphagous fly species, and successfully developed a WeChat Mini Program based on the model. The results show that the model can achieve fast and effective identification of ten common sarcosaphagous flies in Hainan, and the overall correct rate reaches 94.00%. For the few cases of identification difficulties and suspicious results, we have also constructed a rapid molecular species identification system based on DNA Barcoding technology to achieve accurate species identification of the flies under study. As the local fly database continues to be improved, the model is expected to be applicable to local forensic practice.