Method and device for predicting key points of tested object, equipment and storage medium

Abstract

The invention provides a method and device for predicting key points of a tested object, electronic equipment and a storage medium. The method comprises the steps: obtaining a plurality of annotationsets based on key point definition annotations, the annotation sets being used for recording coordinate parameters of one annotation person for annotating each key point of the tested object; calculating the correlation among the coordinate parameters of the key points according to the obtained annotation set; and determining whether to update the key point definition or not according to the calculated correlation. Therefore, the definiteness of key point definition is enhanced; the deviation of understanding of different annotation personnel on the same key point definition is reduced; the availability of the key points is improved; and an important foundation stone is laid for improving the development efficiency of a deep learning project and the model quality.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method, device, equipment and storage medium for predicting key points of a measured object. Background technique [0002] In supervised deep learning, it is usually necessary to label the data. For example, for the key point detection task, the data label is the key point coordinates of the measured object in the image. Among them, the definition and selection of key points are usually project-oriented. Taking human body key point detection as an example, the currently more commonly used COCO model includes 18 human body key points. These key points include points on the main joints of the human body. , such as points on joints such as elbows, wrists, knees, and ankles. There are two main reasons for choosing these key points as detection targets: first, the position information of such key points is very important for judging the posture and movement of the ...

AI Technical Summary

Problems solved by technology

But for most new projects, people often need to redefine the key points according to the project's goals. This process of defining key points is usually subjective, so it is likely that the detected key points will occur because the key points are not clearly defined. The problem of poor usability of points

Moreover, such problems are usually discovered after the data preparation is completed, or even after the data training is completed, which will bring huge time and economic losses to project development

Images