Action positioning method and device, electronic equipment and storage medium

Abstract

The invention provides an action positioning method and device, electronic equipment and a storage medium. The method comprises the following steps: determining a to-be-positioned video; based on an action positioning model, selecting related frames of each action type from the video, performing action classification on each image frame in the video, and determining a frame-level action positioning result based on the related frames of each action type and the action classification result of each image frame, wherein the action positioning model is obtained by training on the basis of a sample video and a video-level action label of the sample video. According to the method, the device, the electronic equipment and the storage medium provided by the invention, the weak supervision training of the action positioning model is realized by applying the video-level action label, and the frame-level action label does not need to be labeled in advance, so that the workload of manual labeling is greatly reduced, the acquisition difficulty of the training sample is reduced, and the availability of action positioning is favorably improved. And the action positioning model performs analysis and prediction from two frame level directions, so that the reliability and accuracy of action positioning under weak supervision can be further improved.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to an action positioning method, device, electronic equipment and storage medium. Background technique [0002] Action localization for video is one of the key issues in video understanding, and it has important applications in the fields of human-computer interaction and human action understanding. [0003] The current action positioning is mostly realized based on the deep learning model, and the deep learning model needs to manually label the action labels of each image frame in the sample video during the training process. Frame-level label labeling consumes a lot of manpower and time costs, and is difficult to obtain. It is huge, and the action displayed by a single frame image is often not clear, resulting in poor accuracy of the frame-level labels obtained by labeling. Contents of the invention [0004] The present invention provides an action localization method...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides an action positioning method, device, electronic equipment and storage medium, which are used to solve the problems that the existing action positioning method requires a large number of frame-level labels during supervised learning, which is time-consuming, laborious and has poor reliability.

Images