Antimicrobial resistance is an emerging global health threat, with integrons playing a major role in the acquisition, expression, and transmission of antibiotic resistance genes. Rapid and accurate identification and classification of integrons would greatly aid clinical practice and environmental surveillance of pathogens. In this study, a comprehensive integron integrease (IntI) sequence database (INTNet-DB) was built, expanding on IntI in INTGRALL by over eightfold. The evolutionary history of IntI was analyzed, and a new systematic IntI classification system was developed based on the phylogeny. Utilizing INTNet-DB, a multi-task multilabel deep learning program, INTNet, was designed and developed to accurately identify and classify integron integrases, their bacterial hosts, and potentially associated ARGs based on the genetic sequences. INTNet is the first deep learning model and demonstrated high performance, achieving precision over 90%, recall above 91%, and accuracy exceeding 94% across validation tests. INTNet is freely available at https://github.com/id-bioinfo/INTNet, and provides a valuable tool for clinical and environmental genomic surveillance.