.png){kind=icon}
.png){kind=icon}
Return
Dilated Multi-scale Fusion for Point Cloud Classification and Segmentation
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-021-11825-9.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
We propose a novel network called Dilated Multi-Scale Fusion network (DMSF) for point cloud analysis in this paper. The network aims to integrate different scales to enhance the feature of point cloud, and each scale feature is obtained by Dilated K - Nearest Neighbor (DKNN) operation, which significantly enhances the size of the receptive field of point cloud. Experimental results show that compared with other state-of-art methods, the proposed network can obtain comparable or even better results in the representative public dataset for point cloud classification and segmentation tasks. Specifically, the oAcc reached 93.6 on the ModelNet40 classification dataset, and the mIoU reached 67.2 on the S3DIS segmentation dataset.
Keywords:
Point cloud feature
Classification
Segmentation
Dilated KNN
Multi-scale fusion
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W