Facial beauty prediction method and device based on multi-task migration

A facial beauty prediction device based on multi-task migration, comprising:

• a similarity measurement module, configured to perform similarity measurement based on a graph structure on a plurality of tasks to obtain an optimal combination of the plurality of tasks, wherein the plurality of tasks comprise a main task for facial beauty prediction and several auxiliary tasks for recognition of facial beauty factors;
• a model construction module, configured to construct a facial beauty prediction model based on the optimal combination, wherein the facial beauty prediction model comprises a feature sharing layer for extracting shared image features of the plurality of tasks;
• a feature parameter migration module, configured to migrate feature parameters of an existing large-scale facial image network to the feature sharing layer of the facial beauty prediction model;
• a pre-training module, configured to input facial images for training to pre-train the facial beauty prediction model;
• and
• a result obtaining module, configured to input a facial image to be tested to the trained facial beauty prediction model to obtain facial recognition results.