Self-adaptive enhanced adversarial hash unsupervised cross-modal retrieval method

The invention relates to the technical field of unsupervised cross-modal Hash retrieval, in particular to an adaptive enhanced adversarial Hash unsupervised cross-modal retrieval method, which comprises the following steps of: respectively carrying out feature extraction on image data and text data to obtain original image features and original text features; designing a text generator to fit text data distribution of original text features, randomly generating false data consistent with image feature dimensions, finally training an image discriminator by using the false data, and generating a feature matrix; mapping the design feature matrix into a Hash code matrix, and quantifying the loss of information in the mapping process to obtain an information amount loss matrix; the method comprises the following steps of: embedding an information amount loss matrix into the construction of a multi-modal similarity matrix to obtain a similarity measurement matrix, realizing semantic alignment of original feature sp