Point cloud data segmentation model training method and device, storage medium and electronic equipment

Abstract

The invention provides a point cloud data segmentation model training method and device, a storage medium and electronic equipment, and the method comprises the steps: obtaining a labeling result of an initial model for first point cloud data, the first point cloud data being point cloud data which is not manually labeled; determining sample point cloud data used as a training sample from the first point cloud data according to the labeling result; and performing weighted training on the initial model based on the sample point cloud data to converge the initial model, the weight of each point of the sample point cloud data being negatively correlated with the corresponding confidence. In the application, the sample point cloud data used as the training sample is labeled by the initial model, and a new training sample can be obtained without manual labeling, so that the problems of high difficulty, long period, high cost and the like of manual labeling of the three-dimensional point cloud can be solved.

Description

technical field [0001] The present application relates to the field of machine learning, in particular, to a point cloud data segmentation model training method, device, storage medium and electronic equipment. Background technique [0002] With the application of various automation equipment in agriculture more and more widely, automation equipment can provide more and more information, including two-dimensional image information and three-dimensional point cloud information. The task (or process) of 3D point cloud segmentation is an important step in the application process of 3D point cloud information. The task of 3D point cloud segmentation relies more on the annotation data of 3D point cloud data. [0003] Compared with the labeling of 2D images, the labeling of 3D point cloud data is more difficult. At present, the demand for 3D point cloud segmentation tasks is increasing. The existing 3D point cloud segmentation algorithms rely heavily on labeling data, but the la...

AI Technical Summary

Problems solved by technology

At present, the demand for 3D point cloud segmentation tasks is increasing. The existing 3D point cloud segmentation algorithms rely heavily on labeling data, but the labeling of 3D point cloud data is difficult, long-term, and costly.

As a result, the realization of 3D point cloud segmentation is difficult, long-term and costly

Images