Brain tumor classification method based on OfficientNetV2 network

The invention provides a brain tumor classification method based on an OfficientNetV2 network. According to the method, the brain tumor classification accuracy and classification precision can be improved. According to the method, an EficientNetV2 network structure is introduced, the network has the characteristics of less parameter quantity, higher convergence speed and higher precision, an improved EficientNetV2 network model introduces a CA lightweight attention mechanism module, and the module can help the network to more accurately position target information, so that an interested area can be more accurately captured, the recognition capability is enhanced, and the recognition efficiency is improved. The problems of low classification accuracy and overfitting of brain tumor MRI images are solved. And then the activation function replaces the Hard-Swsh, so that the numerical stability of the network is improved, and the calculation speed is increased. According to the method, the classification accuracy