Feature processing method and device, electronic equipment and computer readable storage medium

Abstract

The invention provides a feature processing method and device, electronic equipment and a computer readable storage medium, and belongs to the technical field of computers. The method comprises the steps of: performing at least three times of grouping convolution operation on any to-be-processed feature, and obtaining at least three feature components, wherein the to-be-processed feature comprisesframe features of a target video frame in a to-be-processed video, at least three times of grouping convolution operation are carried out on the to-be-processed feature, and at least three feature components are obtained; and calculating an attention graph based on a self-attention mechanism based on a first feature component and a second feature component in the at least three feature components, and obtaining video features of the to-be-processed video according to the remaining feature components, the attention graph, each to-be-processed feature and a preset weight. The method comprises agrade grouping structure, and convolution operation is carried out by adopting grouping convolution operation, so that parameter redundancy and calculation cost can be reduced to a certain extent.

Description

technical field [0001] The invention belongs to the technical field of computers, and in particular relates to a feature processing method, device, electronic equipment and computer-readable storage medium. Background technique [0002] In video understanding technology, it is often necessary to extract frame features of some video frames in the video to be processed, and then gather these frame features into video-level video features. [0003] In the prior art, frame feature aggregation can be performed through a non-local block. Since the frame feature aggregation is performed directly on the entire feature to be processed, the convolution operation, therefore, in this processing method, there will be problems of high parameter redundancy and high calculation cost. Contents of the invention [0004] The purpose of the embodiments of the present invention is to provide a feature processing method, device, electronic equipment, and computer-readable storage medium to sol...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the embodiments of the present invention is to provide a feature processing method, device, electronic equipment, and computer-readable storage medium to solve the problems of high parameter redundancy and high calculation cost when generating video features

Images