Brain tumor classification method based on improved OfficientNetV2 network

The invention provides a brain tumor image classification method based on an improved OfficientNetV2 network and an attention mechanism, the brain tumor image classification method based on an OfficientNetV2 network model is adopted, a residual structure is added behind a first layer of 3 * 3 convolution layer of a basic OfficientNetV2 network structure, three layers of 3 * 3 convolution layers and a layer of ECA attention mechanism are added in the residual structure, and a brain tumor image classification model is constructed. And a 1 * 1 convolution layer for extracting feature information. Meanwhile, a CMAB attention mechanism is added to short cut connection of Fused-MBConv modules of the first three layers, so that the network more pays attention to key information in a shallow network, finally, an output result of a residual structure and an output result of an MBConv module of the last layer of a backbone network are added, feature information of the shallow layer and feature information of the deep l