Real-time face key point quality evaluation method based on deep learning

• Main Authors: XUE CAN, CAI SHENGSHENG, XU DONGYANG, LUO FURONG, WANG ZHI
• Format: Patent
• Language: Chinese; English
• Published: 09.05.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; PHYSICS

The invention discloses a real-time face key point quality evaluation method based on deep learning. The method comprises the following steps: modifying a face key point detection model structure based on a convolutional neural network; designing a loss function in model training; setting a cut-off use threshold value of a confidence coefficient loss function in model training; finally, the sum of the face key point loss function with the weight and the confidence coefficient loss function is a total loss function; and evaluating the face key point quality in a model reasoning stage. According to the face key point quality evaluation method, face key point positioning and face key point quality evaluation work are completed through one model, face key point quality evaluation is achieved with low computing power, and the face key point quality evaluation method is more real-time and has important engineering significance. 本发明公开一种基于深度学习的实时人脸关键点质量评估方法，包括以下步骤：修改基于卷积神经网络的人脸关键点检测模型结构；在模型训练中设计损失函数；在模型训练中设置置信度损失函数的截