Vector database-based large-scale rapid face image similarity retrieval method

The invention discloses a large-scale rapid face image similarity retrieval method based on a vector database, and the method comprises the following steps: S1, recognizing a face through a face detection and segmentation model: carrying out the face detection and segmentation of an image uploaded by a user through a client program, and obtaining an image only containing a face region, namely a face image for short; and S2, obtaining a face feature vector by using a deep neural adversarial model: obtaining the face feature vector after operation by using the segmented image obtained in the step S1 and using the deep neural adversarial model. According to the method, the face image information data is stored in a minimized and vectorized manner, so that storage resources are greatly saved; more network bandwidth is saved; the response time of the system is prolonged; the flexibility and expansibility of the system are improved; enhancing the relation between the image and the service, and keeping the coupling