3D Shape Classification——Point-based Discrete Convolution Network
点云分类——离散卷积 Pointwise CNN PointCNN 卷积核权重不是分配给点云，而是分配给网格，落在同一网格中的点获得相同的权重。 代表1： Pointwise CNN 卷积核在整个点云空间滑动，逐点进行卷积。对于落在同一格子的点，取他们的平均值与w相乘。关键在于如何分配w。 上图把中心点附近的区域分成栅格，每个栅格内的特征先取平均，然后再乘以栅格内的卷积权重，得到一个栅格的
相关文章
- 1. 3D Shape Classification——Point-based Continuous Convolution Networks
- 2. 3D Shape Classification——Point-based Graph based
- 3. 论文笔记《Active Convolution: Learning the Shape of Convolution for Image Classification》
- 4. 【论文笔记】Active Convolution: Learning the Shape of Convolution for Image Classification
- 5. 3D Shape Classification——Point-based Data Indexing based
- 6. Graphic Convolution Network(GCN)
- 7. Paper reading：PointGrid： A Deep Network for 3D Shape Understanding
- 8. 3D Machine Learning Single Object Classification
- 9. convolution backbone network——mobilenetv3
- 10. Transposed convolution & Fully Convolutional Neural Network
- 更多相关文章...
- • Lua 面向对象 - Lua 教程
- • C# 多态性 - C#教程
- • 互联网组织的未来：剖析GitHub员工的任性之源
- • Flink 数据传输及反压详解
相关标签/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
最新文章
欢迎关注本站公众号,获取更多信息
相关文章
- 1. 3D Shape Classification——Point-based Continuous Convolution Networks
- 2. 3D Shape Classification——Point-based Graph based
- 3. 论文笔记《Active Convolution: Learning the Shape of Convolution for Image Classification》
- 4. 【论文笔记】Active Convolution: Learning the Shape of Convolution for Image Classification
- 5. 3D Shape Classification——Point-based Data Indexing based
- 6. Graphic Convolution Network(GCN)
- 7. Paper reading：PointGrid： A Deep Network for 3D Shape Understanding
- 8. 3D Machine Learning Single Object Classification
- 9. convolution backbone network——mobilenetv3
- 10. Transposed convolution & Fully Convolutional Neural Network