Serialized face key point detection method with relay supervision based on deep learning

Abstract

The invention discloses a serialized face key point detection method with relay supervision based on deep learning, and the method comprises the following steps: giving a detected face image, designing a cascade network structure composed of a plurality of modules, and achieving the serialized prediction of key points through the structure through the relay supervision of each module; calculatingthe expectation of the key point coordinates on the output key point heat map by combining the soft maximum value function, completing the conversion from heat map regression to numerical value coordinate regression based on the key point coordinates. The respective advantages of the two regression methods are brought into full play. According to the method, on one hand, the defect that a heat mapregression method cannot perform end-to-end training is overcome, on the other hand, an additional data processing process is omitted, the model training process is accelerated, and the algorithm efficiency is improved. When numerical coordinate regression is carried out, a new segmented loss function is adopted, and finally high positioning accuracy is obtained.

Description

technical field [0001] The invention relates to the technical field of image processing and pattern recognition, in particular to a serialized human face key point detection method with relay supervision based on deep learning. Background technique [0002] Face keypoint detection is one of the classic problems in computer vision, and it is also an important part of the face analysis process. Based on accurate key point detection results, applications such as face pose estimation, expression analysis, and beauty effects can be realized. [0003] The current mainstream face key point detection methods are mainly divided into two categories, namely regression-based methods and convolutional neural network-based methods. The regression-based method is to directly learn the mapping from the appearance of the image to the position of the key point, the most representative of which is the cascade regression, which adopts a strategy from coarse to fine, and the shape increment Δs ...

AI Technical Summary

Problems solved by technology

The numerical coordinate regression method generally has low model complexity and fast speed, but it often has poor performance when there are a large number of key points; the key point detection algorithm based on the heat map implicitly models the key points by introducing the key point heat map. Spatial position relationship, so its spatial generalization ability is strong, but because the output feature map is large, this method obtains performance improvement at the expense of a certain speed and increasing the complexity of the model

Moreover, this method requires additional data preprocessing and postprocessing, and cannot be trained end-to-end

Images