SegGCN: Efficient 3D Point Cloud Segmentation with Fuzzy Spherical Kernel
CVPR 2020
主要是由2019CVPR
Octree guided cnn with spherical kernels for 3d point clouds.
上继承发展,两篇文章作者完全一致。
摘要
原文 | 译文 |
---|---|
Fuzzy clustering is known to perform well in real-world applications. | 模糊聚类在实际应用中表现出色 |
Inspired by this observation, we incorporate a fuzzy mechanism into discrete convolutional kernels for 3D point clouds as our first major contribution. | 基于这个现象,本文的第一个创新点是在3将模糊机制与3D点云中的离散卷积核相结合。 |
The proposed fuzzy kernel is defined over a spherical volume that uses discrete bins. Discrete volumetric division can normally make a kernel vulnerable to boundary effects during learning as well as point density during inference. However, the proposed kernel remains robust to boundary conditions and point density due to the fuzzy mechanism. | 本文提出的模糊核是定义在离散块上的spherical网格上。体素离散化使得卷积对于边界和密度不够鲁棒,但是本文提出的模糊核很好的解决了边界和密度分布不均的影响。 |
Our second major contribution comes as the proposal of an efficient graph convolutional network, SegGCN for segmenting point clouds. | 本文的第二个贡献在于提出了一个高效的图卷积网络结构,适用于点云分割的SegGCN |
The proposed network exploits ResNet like blocks in the encoder and 1 × 1 convolutions in the decoder. SegGCN capitalizes on the separable convolution operation of the proposed fuzzy kernel for efficiency. | 本文提出的网络利用该了ResNet的结构,并且利用提出的可分离的模糊核卷积,提高了效率。 |
We establish the effectiveness of the SegGCN with the proposed kernel on the challenging S3DIS and ScanNet real-world datasets. Our experiments demonstrate that the proposed network can segment over one million points per second with highly competitive performance. | 我们在S3DIS和ScanNet上做了实验,验证了本文提出方法的有效性。 |
Fuzzy kernel
2019 CVPR那篇文章存在的主要问题在于根据sphercial进行点云划分的时候,对于边界上的点处理不好,文章的示意图如下
对于中心点i,s和t的影响应该是差不多的,在左图中,受到划分边界的影响,s和t对于的卷积权重是不同的,这是不理想的。作者针对这一现象,进行了该井,提出了右图的fuzzy kernel。
SegGCN
作者提出的SegGCN网络结构如下图所示
实验部分
首先,作者在S3DIS和ScanNet两个场景分割数据集上给出了mIoU。
然后,作者给出了discussio
kernel compare
主要和19年ICCV KPConv进行了对比
Robustness to point density (missing data)
给出了在不同missing data ratio下的mIoU和mAcc
Decoder module choice
分析了2种decoder模块,本文选用的是1X1的卷积,对比是的2个resnet 模块。说明了用1x1的卷积核可以又好又快。