Image processing method, device and equipment and computer readable storage medium

Abstract

The embodiment of the invention discloses an image processing method, device and equipment and a computer readable storage medium, and the method comprises the steps of obtaining an image, inputting the image into an image segmentation model, and generating at least two image feature matrixes of the image, wherein the image segmentation model comprises a convolution splicing module, the convolution splicing module comprises at least two point-by-point convolution kernels and at least two depth-by-depth convolution kernels, and wherein the features output by the convolution splicing module havea target channel number; performing convolution on the at least two image feature matrixes through each point-by-point convolution kernel, so that at least two intermediate feature matrixes can be generated; according to the at least two depth-by-depth convolution kernels, respectively carrying out convolution on the at least two intermediate feature matrixes to generate at least two feature matrixes to be spliced; and performing feature splicing processing on the at least two intermediate feature matrixes and the at least two to-be-spliced feature matrixes to obtain at least two target feature matrixes. By adopting the invention, the parameter quantity of the model can be reduced, so that the model running speed can be increased.

Description

technical field [0001] The present application relates to the technical field of the Internet, and in particular to an image processing method, device, equipment, and computer-readable storage medium. Background technique [0002] With the rapid popularization of deep learning technology and the improvement of computing power, the performance of semantic segmentation technology has been greatly improved. [0003] Portrait segmentation is one of the basic topics of semantic segmentation, and has received extensive attention in both academia and industry. In applications such as video conferencing and live chat scenes, the functions of these applications can usually be enriched by the portrait segmentation technology. However, the portrait segmentation technology based on deep learning has the problem of a large number of parameters, which leads to a large amount of model calculation. If these applications are run If the hardware capability of the terminal is not high enough,...

AI Technical Summary

Problems solved by technology

In applications such as video conferencing and live chat scenes, the functions of these applications can usually be enriched by the portrait segmentation technology. However, the portrait segmentation technology based on deep learning has the problem of a large number of parameters, which leads to a large amount of model calculation. If these applications are run If the terminal hardware capability is not high enough, it will lead to slow running speed and high resource consumption, and the terminal may not be able to run applications with portrait segmentation technology

Images