面向点云理解的双邻域图卷积方法
|
李宗民,徐畅,白云,鲜世洋,戎光彩
|
Dual-neighborhood graph convolution method for point cloud understanding
|
Zongmin LI,Chang XU,Yun BAI,Shiyang XIAN,Guangcai RONG
|
|
表 3 在ShapeNetPart基准上测试的分割结果 |
Tab.3 Segmentation result on ShapeNetPart benchmark |
|
方法 | 年份 | mIoUcls/% | mIoUins/% | PointNet[7] | 2017 | 80.4 | 83.7 | PointNet++[8] | 2017 | 81.9 | 85.1 | KPConv[18] | 2019 | 85.0 | 86.2 | DGCNN[13] | 2019 | 82.3 | 85.2 | PointASNL[11] | 2020 | — | 86.1 | PAConv[19] | 2021 | 84.6 | 86.1 | CurveNet[41] | 2021 | — | 86.6 | PointTransformer[31] | 2020 | 83.7 | 86.6 | StratifiedFormer[16] | 2022 | 85.1 | 86.6 | PointMLP[46] | 2022 | 84.6 | 86.1 | Point2vec[47] | 2023 | 84.6 | 86.3 | MKConv[45] | 2023 | — | 86.5 | DNG-Net(本文方法) | 2024 | 84.7 | 86.7 |
|
|
|