Unlike traditional machine learning methods, the CNNCF was trained in an end-to-end manner, which ensured the flexibility of the framework for different data sets without much adjustment.