Image feature extraction method and face feature generation equipment

Abstract

The invention discloses a method for determining output features of a target point based on a neighborhood point set of the target point, an image feature extraction method, face feature generation equipment, electronic equipment and a computer readable storage medium. The method comprises the steps: obtaining the neighborhood point set of the target point, and obtaining a neighborhood feature setof the neighborhood point set, wherein the neighborhood point set comprises one or more neighborhood points in a neighborhood with the target point as the center, and the neighborhood feature set comprises input features of the target point and input features of the one or more neighborhood points; dynamically calculating a parameter matrix of the neighborhood feature set based on the input features of the target point and the neighborhood feature set; and generating an output feature of the target point based on the neighborhood feature set and the parameter matrix. Compared with the traditional convolution, the method for extracting the local features has the advantage that the output features can be calculated more lightly and efficiently.

Description

technical field [0001] The present disclosure relates to the field of artificial intelligence, and more specifically relates to a method for determining an output feature of a target point based on a neighborhood point set of the target point, an image feature extraction method, a facial feature generation device, an electronic device, and a computer-readable storage medium. Background technique [0002] At present, most artificial intelligence neural networks involve convolution operations, especially those applied in the field of computer vision. Computer vision is a science that studies how to make machines "see". More specifically, it refers to the use of cameras and computers instead of human eyes to identify, track and measure targets, and further graphics processing, so that computer processing It becomes an image that is more suitable for human eyes to observe or sent to the instrument for detection. For example, neural networks involved in the field of computer vis...

AI Technical Summary

Problems solved by technology

[0003] In order to fuse more information to obtain output features that can more accurately characterize image input, video input, voice input, etc., convolution operations in neural networks usually design larger convolution kernels, or superimpose multiple convolution operations, As a result, a large number of parameters need to be trained in the convolutional layer of the neural network, and a large number of calculations need to be performed during the reasoning process.

Images