Face recognition method and device, electronic equipment and storage medium

Abstract

The invention discloses a face recognition method and device, electronic equipment and a computer readable storage medium. The method comprises the steps of obtaining a training set, wherein the training set comprises a picture of a face wearing a mask; determining a sampling category number and a sampling coefficient of each training, wherein the sampling coefficient represents the sampling number of each category in each training; in each training process, according to the number of the sampling categories, selecting a target category needing to be sampled in the training, and according to the sampling coefficient, selecting a training sample of the training under each target category; and training a face recognition model based on the training sample so as to perform face recognition by using the trained face recognition model. Therefore, according to the face recognition method provided by the invention, the accuracy of the face recognition model is improved.

Description

technical field [0001] The present application relates to the technical field of face recognition, and more specifically, to a face recognition method and device, an electronic device, and a computer-readable storage medium. Background technique [0002] Face recognition is a typical open-set task, that is, the object during training is not the same as the object during actual use. Face recognition needs to be applied in extremely harsh scenarios, such as when masks and hats block a large area of ​​the face. The number of training samples in such scenarios is much lower than that in ordinary scenarios. In the existing technology, when training the face recognition model, the particularity of the samples is often not distinguished, and a large number of face samples such as normal illumination, high resolution, no mask, no occlusion, etc. are mixed with a small number of face samples wearing masks, severe occlusion, etc. Special scene face samples are randomly sampled for tr...

AI Technical Summary

Problems solved by technology

For the training set with an unbalanced number of samples, the above scheme can easily cause a long-tail problem, that is, the training of the category with a large number of samples is more sufficient, while the training of the category with a small number of samples is not sufficient, so that the face obtained in the final training The robustness of the recognition model is poor, that is, the face target recognition effect in special situations such as wearing a mask and large area occlusion is not good

It can be seen that the above scheme is likely to cause insufficient training of face picture categories with a small number of samples, resulting in low accuracy of the face recognition model

Images