Fundus photo automatic detection method based on image recognition and system thereof

The invention discloses a fundus photo automatic detection method based on image recognition. The method comprises the following steps: firstly, obtaining fundus photos and preprocessing; secondly, dividing an eye fundus photo region, and performing region depth feature extraction; generating a global feature vector; and finally, training the classification network through the global feature vector, and completing eye fundus photo classification detection through the trained classification network. According to the method, the number of network parameters is effectively reduced, and the overfitting problem of the model is relieved; and the robustness of translation and deformation is obtained. The regional and global features are obtained through partitioning and merging, so that more feature information can be explored. A global pooling layer is used for replacing a full connection layer to fuse learned depth features, so that redundancy of the network is effectively avoided, and the prediction performance i