A face key point detection method based on lookup table activation function

Abstract

The invention discloses a face key point detection method based on a lookup table activation function, which belongs to the technical field of image detection, specifically proposes a lookup table based activation function realization method, and applies it to face key point detection, which can Greatly improve the nonlinear expression ability of the network, improve the accuracy of face key point detection, and the calculation amount is small; the face area re-determined by the full convolutional network can effectively reduce the impact of face key point detection on the original face frame. Dependence, no matter where the actual face is located in the corner of the original face frame, the algorithm in this paper can accurately detect the coordinates of the key points of the face; the network training introduces the LUT lookup table activation function to fit the complex mapping function and increase the neural network. Non-linear expression ability; LUT lookup table activation function calculation is simple and fast.

Description

technical field [0001] The invention relates to the technical field of image detection, in particular to a face key point detection method based on a lookup table activation function. Background technique [0002] Face key point detection, also known as face key point detection, positioning or face alignment, refers to locating the key areas of the face for a given face image, including eyebrows, eyes, nose, mouth, face contour, etc. . Face key point detection methods are roughly divided into three types, which are traditional methods based on ASM (Active Shape Model) and AAM (Active Appearnce Model), methods based on cascaded shape regression, and methods based on deep learning. [0003] Deep learning has developed rapidly in recent years. Represented by neural networks, it has solved problems that were difficult to solve in many fields before. The basic consensus currently formed is to add an activation function at the end of each layer of the neural network to improve t...

AI Technical Summary

Problems solved by technology

The basic consensus currently formed is to add an activation function at the end of each layer of the neural network to improve the nonlinear expression ability of the neural network. The commonly used activation function is relu, and some variants include prelu, leaky_relu, etc. These activation functions can be used to a certain extent. Increase the network nonlinearity, but the form is too fixed and not flexible enough

[0004] In face key point detection, the nonlinear expression ability of the currently used neural network is poor, resulting in low accuracy and high computational complexity of face key point detection

Images