What is it about?
This work improves the performance of neural networks on spherical images and 3D meshes. It does so through a interpolated convolution that operates across of the whole surface of the object. This method improves performance on other tasks such as segmentation, but it specifically is showcased for spherical and 3D style transfer.
Featured Image
Photo by Pawel Czerwinski on Unsplash
Why is it important?
Consistent style transfer for spherical images and 3D meshes could allow for unique stylistic choices in the 3d graphics pipeline.
Read the Original
This page is a summary of: Interpolated SelectionConv for Spherical Images and Surfaces, January 2023, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/wacv56688.2023.00040.
You can read the full text:
Contributors
The following have contributed to this page
