Identity recognition model training method and device, police scooter and storage medium

Abstract

The invention provides an identity recognition model training method, and the method comprises the steps of obtaining a collected image containing a human face in an actual scene as a training image, extracting features in the training image, and carrying out the extraction to obtain human face features and non-human face features; taking the non-face feature as the input of a non-face recognition model in the identity recognition model, and obtaining an output first identity recognition result; taking the face feature as the input of a face recognition model in the identity recognition model, and obtaining an output second identity recognition result; comparing the first identity recognition result with the second identity recognition result, and if the first identity recognition result is consistent with the second identity recognition result, improving the pre-judgment weight of a non-face recognition model; and when the pre-judgment weight is greater than a preset threshold value, judging that the training of the non-face recognition model is completed. According to the method, model training can be conveniently carried out, and the trained model can rapidly and accurately carry out identity recognition. In addition, the invention further provides the identity recognition model training device, a police scooter and a storage medium.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to a training method, device, police scooter and storage medium for an identity recognition model. Background technique [0002] At present, face recognition technology is relatively mature, and the accuracy of detection and recognition is relatively high in a standard environment, especially the accuracy of face recognition based on fixed equipment is relatively high, such as face recognition in gates, access control, etc. . However, face recognition is often poor in extreme environments. For example, face recognition in strong light, rainy, foggy and other environments is often inaccurate. [0003] Furthermore, for police officers, when carrying out police work such as investigation, search, and arrest, face recognition is based on mobile devices, for example, installed on police cars, and it is necessary to realize target recognition during movement , the face image...

AI Technical Summary

Problems solved by technology

However, face recognition is often poor in extreme environments, for example, face recognition is often inaccurate in strong light, rainy, foggy and other environments

[0003] Furthermore, for police officers, when carrying out police work such as investigation, search, and arrest, face recognition is based on mobile devices, for example, installed on police cars, and it is necessary to realize target recognition during movement , the face images captured during the movement are often blurred, resulting in relatively low recognition accuracy

[0004] In order to solve the above problems, the existing methods usually design a more complex model structure, but the more complex the model structure, the slower the recognition speed will be, which cannot meet the recognition speed requirements of the police, and the training of the model often requires a lot of training data, and the training data often needs to be artificially standardized, resulting in not only time-consuming and labor-intensive model training, but also the failure to achieve the purpose of fast and accurate recognition

Images