What is it about?
This work shows how to take a neural network that was trained on regular 2D images and use them on other data types like spherical images or even 3D data. It does so using specialized graph techniques. The results are shown for task such as style transfer and depth prediction.
Featured Image
Photo by Lok Yiu Cheung on Unsplash
Why is it important?
This work makes artificial intelligence techniques more accessible to new data domains, leading to less specialized networks. This saves on training time, needed data, and computational resources.
Read the Original
This page is a summary of: SelectionConv: Convolutional Neural Networks for Non-rectilinear Image Data, January 2022, Springer Science + Business Media,
DOI: 10.1007/978-3-031-20071-7_19.
You can read the full text:
Contributors
The following have contributed to this page
