Semi-supervised face feature point detection result evaluation method and device and storage medium

Abstract

The invention provides a semi-supervised face feature point detection result evaluation method and device and a storage medium, an evaluation result for evaluating the accuracy of a current prediction result is added on the basis of a traditional deep learning-based face detection algorithm, so that an algorithm model can evaluate the prediction quality without manual marking reference, which conditions the trained model is poor in performance can be explored, and targeted improvement can be realized. For example, in an actual application scene, each face image input into a model does not have a face feature point position mark, at the moment, whether the face image is good or not can not be predicted through an artificial label and a prediction result comparison judgment algorithm, however, the method provided by the invention can predict the face feature point coordinates and can additionally provide an evaluation result for evaluating the accuracy of the current prediction result. Therefore, which data prediction effect is good and which prediction effect is poor in the use process of the model can be known, and targeted improvement on an algorithm is facilitated.

Description

technical field [0001] The present invention relates to the technical field of face detection, in particular to a semi-supervised face feature point detection result evaluation method, device and storage medium. Background technique [0002] Facial feature points refer to 5 feature points such as eyes, nose, and 2 corners of the mouth, or 28 feature points including more contour points, or even 64 and 128 facial feature points. The quality of face feature points represents the quality of the face to a certain extent. For example, if the feature points are blocked or blurred, then the face cannot be effectively recognized. In the early stage, the face image with high-quality feature points is selected first. Face recognition can not only improve the accuracy of face recognition, but also improve the efficiency of face recognition. [0003] The face feature point detection algorithm is an algorithm for detecting the position of face feature points, which is widely used in the...

AI Technical Summary

Problems solved by technology

[0004] However, when using the trained model in an actual application scenario, since the input face image at this time does not have a well-marked artificial label, it is impossible to evaluate the accuracy of the model's prediction results.

This makes the prediction results of the face feature point detection model uncertain, which will affect the stability of the entire algorithm system

At the same time, due to the lack of such means of evaluating the accuracy of prediction results, the model does not know which images in application scenarios are more difficult to predict (difficult samples), and thus does not know how to improve the model

Images