Transform-based double-branch face image key point detection method

The invention discloses a dual-branch face image key point detection method based on Transform. The method is specifically implemented according to the following steps: performing feature extraction on a face image by using a dual-branch network; the double-branch network comprises a convolutional neural network, a Transform branch network, a feature fusion module and a progressive refining module; designing a loss function, wherein the loss function comprises Link and Lrefine; training the double-branch network by using the training set of the face database, and storing a trained model; and testing the stored network model, and calculating the normalized error value of the L groups of face key point coordinates Ln finally output by the model and the truth value data. According to the method, the problems that the multi-scale features cannot be extracted and the convolution operation of the convolutional neural network can only capture local information and cannot establish remote dependence of a global image