Video annotation method and device, electronic equipment and computer readable storage medium

Abstract

The embodiments of the invention relate to the technical field of video processing, and disclose a video annotation method and device, electronic equipment and a computer readable storage medium. Thevideo annotation method comprises the steps of: respectively detecting a preset number of videos through at least two detection algorithms on the basis of a preset target detection object; then determining a first annotation result of each video under each detection algorithm through a preset video annotation mode according to a detection result of each detection algorithm for each video; and determining the annotation result of each video as a second annotation result of a predetermined confidence level according to each first annotation result of each video for analysis of video data.

Description

technical field [0001] Embodiments of the present application relate to the technical field of video processing, and specifically, the present application relates to a video tagging method, device, electronic equipment, and computer-readable storage medium. Background technique [0002] With the continuous application of machine learning technology in the field of computer vision, the demand for labeled data is increasing. Due to the time-consuming and labor-intensive shortcomings of the traditional manual data labeling method, it is difficult to meet the needs of computer vision for a large number of labeled data samples, especially for key data in the video (such as the target object included in the video). Labeling is more difficult than simple image data labeling. The labeling of a piece of video data is often equivalent to the labeling of tens of thousands or even millions of image data. [0003] At present, some automated video tagging methods have also appeared. Howe...

AI Technical Summary

Problems solved by technology

Due to the time-consuming and labor-intensive shortcomings of the traditional manual data labeling method, it is difficult to meet the needs of computer vision for a large number of labeled data samples, especially for key data in the video (such as the target object included in the video). Labeling is more difficult than simple image data labeling. The labeling of a piece of video data is often equivalent to the labeling of tens of thousands or even millions of image data

[0003] At present, some automated video tagging methods have also appeared. However, the inventors of the present application found in the implementation process that these automated video tagging methods usually rely on a single detection algorithm to detect the video frame by frame, and then cooperate with the tagging module. The results of frame-by-frame detection are analyzed, and the corresponding video annotations are finally obtained. Since the accuracy of a single detection algorithm is often not 100% accurate, some complex videos will have ambiguous annotations, which seriously affects the credibility of video annotations.

Images