About The Forum
Calendar 2017 Poster Sessions : SyncSpecCNN: Synchronized Spectral CNN for 3D Shape Segmentation
Student Name : Li Yi
Advisor : Leonidas Guibas
Research Areas: Artificial Intelligence, Graphics/HCI
Advisor : Leonidas Guibas
Research Areas: Artificial Intelligence, Graphics/HCI
Abstract:
We study the problem of semantic annotation on 3D models that are represented as meshes. On the shape graphs of meshes, a functional view is taken to represent localized information, so that annotations such as part segment or keypoint are nothing but 0-1 indicator vertex functions. To predict vertex functions by convolutional neural networks, we resort to spectral CNN method that enables weight sharing by parameterizing kernels in the spectral domain defined by graph Laplacian eigenbases. Under this setting, our network, named SyncSpecCNN, strive to overcome two key challenges: how to share coefficients and conduct multi-scale analysis in different parts of the graph for a single shape, and how to share information across related but different shapes that may be represented by very different graphs. Towards these goals, we introduce a spectral parameterization of dilated convolutional kernels and a spectral transformer network. Experimentally we have tested our SyncSpecCNN on various tasks, including 3D shape part segmentation and 3D keypoint prediction. State-of-the-art performance has been achieved on all benchmark datasets.
Bio:
Li Yi is a 4th year PhD student at Stanford University, supervised by Professor Leonidas J. Guibas. Prior to joining Stanford, he received his bachelor degree in Electronic Engineering from Tsinghua University. His research interest lies in the area of 3D computer vision and shape analysis. Specifically, he focuses on data driven 3D object understanding, with potential applications in robots, virtual/augmented reality.
We study the problem of semantic annotation on 3D models that are represented as meshes. On the shape graphs of meshes, a functional view is taken to represent localized information, so that annotations such as part segment or keypoint are nothing but 0-1 indicator vertex functions. To predict vertex functions by convolutional neural networks, we resort to spectral CNN method that enables weight sharing by parameterizing kernels in the spectral domain defined by graph Laplacian eigenbases. Under this setting, our network, named SyncSpecCNN, strive to overcome two key challenges: how to share coefficients and conduct multi-scale analysis in different parts of the graph for a single shape, and how to share information across related but different shapes that may be represented by very different graphs. Towards these goals, we introduce a spectral parameterization of dilated convolutional kernels and a spectral transformer network. Experimentally we have tested our SyncSpecCNN on various tasks, including 3D shape part segmentation and 3D keypoint prediction. State-of-the-art performance has been achieved on all benchmark datasets.
Bio:
Li Yi is a 4th year PhD student at Stanford University, supervised by Professor Leonidas J. Guibas. Prior to joining Stanford, he received his bachelor degree in Electronic Engineering from Tsinghua University. His research interest lies in the area of 3D computer vision and shape analysis. Specifically, he focuses on data driven 3D object understanding, with potential applications in robots, virtual/augmented reality.