Multi-task neural network face key point detection method for edge device

Abstract

The invention relates to the field of deep learning, face recognition and face key point detection, and provides a multi-task neural network face key point detection method for edge equipment, which is used for realizing face key point calibration and face accurate recognition on mobile equipment. Therefore, the technical scheme adopted by the invention is that according to the multi-task neural network face key point detection method for the edge device, a face image needing to be detected is input into a convolutional neural network, and the convolutional neural network outputs detected face key point coordinates; and defining a loss function of the convolutional neural network. The method is mainly applied to face recognition and face key point detection occasions.

Description

technical field [0001] The invention relates to the fields of deep learning, human face recognition and human face key point detection, in particular to a multi-task neural network human face key point detection method for edge devices. Background technique [0002] Face keypoint detection, also known as face calibration or face alignment, aims to automatically locate a set of predefined reference points on the face (eg, eye corners, nose tip, mouth corners, etc.). As a fundamental component of various face applications such as face recognition [1, 2], face verification [3], face morphing [4], and face editing [5], this problem has been receiving much attention from computer vision-related fields. attention, and has made great progress in the past few years. However, limited by factors such as detection accuracy, processing speed, and model size, it is still challenging to develop a practical face keypoint detection technology. [0003] The difficulty of the technology is ...

AI Technical Summary

Problems solved by technology

[0004] 1. Local changes: Face images will be subject to local interference caused by facial expressions, local extreme lighting (such as highlights and shadows), and occlusions, which may cause some key points to be undisplayed or abnormally positioned.

[0005] 2. Global change: Pose and image quality are two key factors, which will have a global impact on the appearance of the face in the image. When the global structure of the face is misestimated, it will lead to incorrect positioning of most key points. precise

[0006] 3. Data imbalance: It is very common for the data sets currently available for training to be unevenly distributed between face categories and attributes.

Imbalance is likely to make the algorithm or model unable to correctly represent the characteristics of the data, thereby reducing the accuracy of detection

[0007] 4. Model efficiency: model size and computing performance also limit the practicality of the algorithm

Although the current deep learning algorithm has made great progress, there are still many shortcomings, especially in practical applications, there is still a lot of room for improvement in the accuracy, efficiency and simplicity of the detection algorithm [28]

Images