Feature construction method, cross-modal retrieval method and device and computer storage medium

The invention provides a feature construction method, a cross-modal retrieval method, a cross-modal retrieval device and a computer storage medium. The feature construction method comprises the following steps: segmenting original modal data to obtain a plurality of segmented data segments; inputting the original modal data into a modal encoder, and extracting local modal features and global modal features of the original modal data; obtaining a target granularity feature of each segmented data segment by using the local modal features; splicing the global modal feature and all the target granularity features to obtain a target modal feature; and establishing an alignment relationship between the target modal feature and a preset query feature to obtain a multi-granularity retrieval feature of the original modal data. According to the feature construction method, the modal data are represented through the local features and the global features, the multi-granularity retrieval features are constructed, and ref