Lightweight depth estimation method and system based on neural network

The invention relates to a lightweight depth estimation method and system based on a neural network. The method comprises the following steps: S1, data preprocessing; s2, building a depth estimation neural network; s3, building a pose estimation network; setting a loss function to train the two networks; and S4, generating a global voxel result according to the output depth estimation and pose estimation results. According to the method provided by the invention, the lightweight depth estimation network is applied to the endoscopic examination image for the first time, and the network combines convolution and attention mechanisms. The method uses a mixture of convolution and attention mechanisms as an encoder to aggregate local texture information and global contour features. According to the method, a competitive result is obtained, and meanwhile, the number of parameters of the depth estimation network is reduced. Compared with a previous method, the proposed attitude network obtains the minimum error on a