Image saliency detection method of double-current decoder based on information complementation

The invention discloses an image saliency detection method of a double-flow decoder based on information complementation. The method comprises the following steps: S1, decomposing a label image to obtain a corresponding main body label image and a contour detail label image; S2, carrying out random cutting, random rotation, normalization and graying processing on the training data set image to enhance the diversity of a sample; S3, inputting an image, pre-processing the image by using a VGG16 framework, and respectively collecting image features of different sizes by using a group of coding blocks of different dimensions; S4, inputting the five-layer output feature map obtained by the encoder into an Embedding layer, and performing dimension unification; S5, respectively transmitting the encoded features of the target main body graph and the encoded features of the contour detail graph into a significance branch and a contour graph branch, and are respectively supervising the branches by using the obtained ma