Video classification method, model training method and device

Abstract

The invention discloses a video classification method, which comprises the following steps that: obtaining an image frame sequence corresponding to a video file to be classified with the image frame sequence comprising T frames of images; obtaining an appearance information feature sequence corresponding to the image frame sequence through an image classification network model, the appearance information feature sequence comprising T appearance information features, and the appearance information features having a corresponding relationship with the image frames; obtaining a motion informationfeature sequence corresponding to the appearance information feature sequence through a motion prediction network model, the motion information feature sequence comprising T motion information features, and the motion prediction network model being used for predicting the motion information features corresponding to the appearance information features; and determining a video classification result of the to-be-classified video file according to the appearance information feature sequence and the motion information feature sequence. The invention further provides a model training method and device. According to the method, the calculation time and resources can be saved, so that the data processing efficiency is improved.

Description

technical field [0001] The present application relates to the field of artificial intelligence, in particular to a video classification method, a model training method and a device. Background technique [0002] Video classification refers to classifying the content contained in a given video segment. Video categories typically include actions (such as making a cake), scenes (such as a beach), and objects (such as a table). Among them, video action classification is the most popular. After all, action itself contains dynamic factors, which cannot be described by static images. Therefore, it is very common to classify videos based on motion features. [0003] At present, the commonly used video classification technology is mainly to first classify the appearance information features extracted from the video red green blue (RGB) image, and then further use the motion information features extracted from the optical flow image to improve the video quality. classification perfo...

AI Technical Summary

Problems solved by technology

[0004] However, in existing solutions, optical flow images need to be calculated based on at least two frames of RGB images. For videos, at a resolution of 24 frames per second, short videos usually contain more than 15 seconds of content, while long videos often contain tens of minutes

Therefore, it takes a lot of time and space to calculate the optical flow image to obtain the optical flow image representation of the video, which leads to low data processing efficiency and consumes more computing resources and storage resources.

Images