eduzhai > Applied Sciences > Engineering >

Wireless 3D Point Cloud Delivery Using Deep Graph Neural Networks

  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: In typical point cloud delivery, a sender uses octree-based digital videocompression to send three-dimensional (3D) points and color attributes overband-limited links. However, the digital-based schemes have an issue called thecliff effect, where the 3D reconstruction quality will be a step function interms of wireless channel quality. To prevent the cliff effect subject tochannel quality fluctuation, we have proposed soft point cloud delivery calledHoloCast. Although the HoloCast realizes graceful quality improvement accordingto wireless channel quality, it requires large communication overheads. In thispaper, we propose a novel scheme for soft point cloud delivery tosimultaneously realize better quality and lower communication overheads. Theproposed scheme introduces an end-to-end deep learning framework based on graphneural network (GNN) to reconstruct high-quality point clouds from itsdistorted observation under wireless fading channels. We demonstrate that theproposed GNN-based scheme can reconstruct clean 3D point cloud with lowoverheads by removing fading and noise effects.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×