A. Notchenko, E. Kapushev, and E. Burnaev. Sparse 3d convolutional neural networks for large-scale shape retrieval. arXiv:1611.09159, 2016.Alexandr Notchenko, Ermek Kapushev, and Evgeny Burnaev. Sparse 3d convolutional neural networks for large-scale shape retrieval. CoRR, abs/1611.09159, ...
Sparse3Dconvolutionalneuralnetworks BenGraham UniversityofWarwick b.graham@warwick.ac.uk May12,2015 Abstract Wehaveimplementedaconvolutionalneuralnetworkdesignedforpro- cessingsparsethree-dimensionalinputdata.Theworldweliveinisthree dimensionalsotherearealargenumberofpotentialapplications.Inthe questforefficiency,we...
[3]:V ote3Deep: Fast Object Detection in 3D Point Clouds using Efficient Convolutional Neural Networks. [4]:Sparse 3D Convolutional Neural Networks. [18]:Octnet: Learning Deep 3D Representations at High Resolutions. [22]:V oting for voting in online point cloud object detection. ...
考虑到LiDAR点云的稀疏性(即大约90%的体素都没有点),Vote3 Deep使用了投票方案(voting scheme),而SECOND使用了3D SpConv和3D SubSpConv来构建基于3D sparse CNN的骨干网络,用于LiDAR点云中的3D目标检测。 3DBN sparse CNN的设计遵循encoder网络架构或encoder/decoder网络架构。Encoder网络架构会使用足够多的连续的3D...
Compressive holography techniques enabled 3D reconstruction from a single 2D hologram for sparse 3D densities32,33. However, for densely packed specimens, decompressive interference using a forward model is an ill-posed problem34. On the contrary, estimating the local depth information and the optical...
5、Interpolation-Aware Padding for 3D Sparse Convolutional Neural Networks(arXiv)https://arxiv.org/pdf/2108.06925 Yu-Qi Yang, Peng-Shuai Wang, Yang Liu 从3D输入中生成稀疏的非空voxels 的Sparse voxel-based CNN 被广泛应用于各种三维视觉任务。论文提出了一种简单而有效的填充方案——插值感知填充，在...
5、Interpolation-Aware Padding for 3D Sparse Convolutional Neural Networks(arXiv) arxiv.org/pdf/2108.0692 Yu-Qi Yang, Peng-Shuai Wang, Yang Liu 从3D输入中生成稀疏的非空voxels 的Sparse voxel-based CNN 被广泛应用于各种三维视觉任务。论文提出了一种简单而有效的填充方案——插值感知填充,在非空voxels...
Advances in neural information processing systems 33 (2020), 6840-6851. 1, 2, 6, 9 [HLA*19] Hu Y., LI T.-M., Anderson L., Ragan-Kelley J., Du-RAND F.: Taichi: a language for high-performance computation on spa-tially sparse data structures. ACM Transactions on Graphics (TOG) ...
在这项工作中,我们选择在ScanNet[54] 上表现最好的SparseConvNet[13] ,作为我们的 3D 网络。 图2:我们用于域适应的跨模态无监督学习架构。有两个独立的网络流:一个 2D 流(红色),它以图像作为输入,并使用 U-Net 风格的 2DConvNet[12];以及一个 3D 流(蓝色),它以点云为输入,并使用 U-Net 风格的 3D...
在这项工作中,我们选择在 ScanNet [54] 上表现最好的 SparseConvNet [13] ,作为我们的 3D 网络。 图2:我们用于域适应的跨模态无监督学习架构。有两个独立的网络流:一个 2D 流(红色),它以图像作为输入,并使用 U-Net 风格的 2D ConvNet [12];以及一个 3D 流(蓝色),它以点云为输入,并使用 U-Net ...