Saved in:
Bibliographic Details
Main Authors: Kundu, Soumyabrata, Kondor, Risi
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.15932
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911230174691328
author Kundu, Soumyabrata
Kondor, Risi
author_facet Kundu, Soumyabrata
Kondor, Risi
contents We introduce Steerable Transformers, an extension of the Vision Transformer mechanism that maintains equivariance to the special Euclidean group $\mathrm{SE}(d)$. We propose an equivariant attention mechanism that operates on features extracted by steerable convolutions. Operating in Fourier space, our network utilizes Fourier space non-linearities. Our experiments in both two and three dimensions show that adding steerable transformer layers to steerable convolutional networks enhances performance.
format Preprint
id arxiv_https___arxiv_org_abs_2405_15932
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Steerable Transformers for Volumetric Data
Kundu, Soumyabrata
Kondor, Risi
Computer Vision and Pattern Recognition
We introduce Steerable Transformers, an extension of the Vision Transformer mechanism that maintains equivariance to the special Euclidean group $\mathrm{SE}(d)$. We propose an equivariant attention mechanism that operates on features extracted by steerable convolutions. Operating in Fourier space, our network utilizes Fourier space non-linearities. Our experiments in both two and three dimensions show that adding steerable transformer layers to steerable convolutional networks enhances performance.
title Steerable Transformers for Volumetric Data
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.15932