Semi-supervised three-dimensional point cloud semantic segmentation method based on neural network

Abstract

The invention belongs to the technical field of deep learning and computer vision, and particularly relates to a semi-supervised three-dimensional point cloud semantic segmentation method based on a neural network. A semi-supervised learning normal form is adopted, a three-dimensional point cloud semantic segmentation network model is combined, and a whole semi-supervised three-dimensional point cloud semantic segmentation method framework is formed; the network model is divided into a student network and a teacher network, and the two networks sample the same SSCNs network; the input of the student network is an original point cloud which is not transformed, and the input of the teacher network is a transformed point cloud; and the output of the labeled part of the student network is supervised by the corresponding label, and the overall output of the student network and the teacher network is subjected to consistency supervision, so that the weight of the student network is updated, and the weight of the teacher network is obtained by performing exponential moving average on the weight of the student network. Experiments show that by using semi-supervised learning with labeled data and unlabeled data, the performance of the network is obviously improved on each labeling rate.

Description

technical field [0001] The invention belongs to the technical fields of deep learning and computer vision, and in particular relates to a method for semantic segmentation of three-dimensional point clouds. Background technique [0002] In recent years, deep learning has achieved outstanding performance on a variety of computer vision tasks, especially in the image field. However, for some practical applications such as autonomous driving, virtual reality, and augmented reality, it is necessary to obtain richer information than pure pictures to achieve better scene understanding. The 3D data collected by lidar or RGB-D depth camera is a good supplement to the 2D image data, and the representation of these 3D data is usually point cloud. 3D point cloud is composed of a large number of points with 3D coordinates and colors. It is an intuitive 3D data format. Compared with 2D images, 3D point cloud contains rich environmental space information, which is more helpful for scene u...

AI Technical Summary

Problems solved by technology

SSCNs [11] first voxelizes the input point cloud, and proposes a new sparse convolution method to alleviate the problem of heavy computing burden of point cloud

Images