Video action positioning method and device

By introducing classification branch sub-model and positioning branch sub-model in the video action positioning model, combining the importance attention mechanism and feature unified sub-module, the conflict between action classification and action positioning in the existing technology is solved, and high accuracy is achieved Sexual video action positioning.

CN114842396BActive Publication Date: 2025-02-21JD DIGITS HAIYI INFORMATION TECHNOLOGY CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202210550538.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-20
Publication Date
2025-02-21
Estimated Expiration
2042-05-20

AI Technical Summary

Technical Problem

The existing video action positioning model needs to take into account both action classification and action positioning, resulting in high purity conflicts between action classification and action positioning, resulting in incomplete positioning or positioning errors.

Method used

By introducing classification branch sub-model and positioning branch sub-model in the video action positioning model, we are responsible for the classification of action types and positioning of action types, and improve the accuracy of feature extraction and processing through the importance attention mechanism and feature unification sub-module.

Benefits of technology

The separation of action classification and positioning is achieved, and the conflict between action classification and action positioning is resolved, so that the classification branch sub-model can be accurately classified, the positioning sub-model can be accurately positioned, and the accuracy of classification in the positioning results is improved by adjusting the results. so as to effectively improve the accuracy of video classification and positioning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114842396B_ABST
    Figure CN114842396B_ABST
Patent Text Reader

Abstract

The present invention discloses a video action location method and device, and relates to the field of computer vision technology. A specific implementation of the method includes: inputting a video sequence of a video to be processed into a pre-trained video action location model; extracting video feature parameters from the video sequence through the video action location model; determining the action type included in the video to be processed based on the video feature parameters and the classification branch sub-model included in the video action location model; locating the action type included in the video to be processed based on the video feature parameters and the location branch sub-model included in the video action location model; adjusting the location result according to the action type determined by the classification branch sub-model. This implementation effectively improves the accuracy of action classification and location.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Video behavior timeline detection method

    CN108830212A