SEARCH SESSIONS

Search All
 
Refine Results:
 
Year(s)

SOCIAL MEDIA

EMAIL SUBSCRIPTION

 
 

GTC On-Demand

Presentation
Media
Abstract:
This presentation shows in-depth comparisons of several neural network models for 3D object classification. Object classification from 2D image is studied thoroughly and widely adopted during last few years by following the advances of deep neural networks. From then, 3D object classification methods are actively studied, and yet not completely mature. Point cloud is most basic format of 3D objects. In this work, we present many neural network models that can be learned from 3D point cloud. It includes directly learning from 3D point cloud, projected 2D pixels, and voxelated volumes. This work uses Princeton ModelNet datasets and ShapeNetCore.v2 dataset, and then provides the comparisons of those neural network models.
This presentation shows in-depth comparisons of several neural network models for 3D object classification. Object classification from 2D image is studied thoroughly and widely adopted during last few years by following the advances of deep neural networks. From then, 3D object classification methods are actively studied, and yet not completely mature. Point cloud is most basic format of 3D objects. In this work, we present many neural network models that can be learned from 3D point cloud. It includes directly learning from 3D point cloud, projected 2D pixels, and voxelated volumes. This work uses Princeton ModelNet datasets and ShapeNetCore.v2 dataset, and then provides the comparisons of those neural network models.  Back
 
Topics:
AI and DL Research, Graphics and AI, Rendering and Ray Tracing, Real-Time Graphics
Type:
Talk
Event:
GTC Silicon Valley
Year:
2018
Session ID:
S8453
Streaming:
Download:
Share:
 
Abstract:

This talk presents sparse voxelization of time-lapse point cloud. Point cloud has several advantages including capturing easiness, data simplicity, and most fundamental 3D primitive. Because of these advantages, the easy way to collect time-lapse 3D information is by capturing point cloud using laser scan or photogrammetry. However, point cloud representation is lack of spatial connectivity and has notoriously big size of captured data. Our sparse volumetric representation fills the gap between the pros and cons of point cloud by keeping the simplicity and easiness and providing spatial connectivity as well as GPU-friendly data structure. In this talk, we show our massive-scale time-lapse point cloud dataset, the compression as sparse voxels, and further processing in parallel and visualization using GVDB in CUDA.

This talk presents sparse voxelization of time-lapse point cloud. Point cloud has several advantages including capturing easiness, data simplicity, and most fundamental 3D primitive. Because of these advantages, the easy way to collect time-lapse 3D information is by capturing point cloud using laser scan or photogrammetry. However, point cloud representation is lack of spatial connectivity and has notoriously big size of captured data. Our sparse volumetric representation fills the gap between the pros and cons of point cloud by keeping the simplicity and easiness and providing spatial connectivity as well as GPU-friendly data structure. In this talk, we show our massive-scale time-lapse point cloud dataset, the compression as sparse voxels, and further processing in parallel and visualization using GVDB in CUDA.

  Back
 
Topics:
Rendering and Ray Tracing, Large Scale and Multi-Display Visualization, Real-Time Graphics
Type:
Talk
Event:
GTC Silicon Valley
Year:
2017
Session ID:
S7108
Download:
Share: