Convolutions on Spherical Images
This work addresses a domain-specific challenge in computer vision for processing spherical images, with incremental improvements in performance.
The paper tackled the problem of applying convolutional neural networks to spherical images by proposing a representation based on the icosahedral Snyder equal-area projection, which improved semantic segmentation results by 12.6% over the state-of-the-art.
Applying convolutional neural networks to spherical images requires particular considerations. We look to the millennia of work on cartographic map projections to provide the tools to define an optimal representation of spherical images for the convolution operation. We propose a representation for deep spherical image inference based on the icosahedral Snyder equal-area (ISEA) projection, a projection onto a geodesic grid, and show that it vastly exceeds the state-of-the-art for convolution on spherical images, improving semantic segmentation results by 12.6%.