We present a simple baseline for directly estimating the relative pose (rotation and translation, including scale) between two images. Deep methods have recently shown strong progress but often require complex or multi-stage architectures. We show that a handful of modifications can be applied to a Vision Transformer (ViT) to bring its computations close to the Eight-Point Algorithm. This inductive bias enables a simple method to be competitive in multiple settings, often substantially improving over the state of the art with strong performance gains in limited data regimes.
@article{arxiv.2208.08988,
title = {The 8-Point Algorithm as an Inductive Bias for Relative Pose Prediction by ViTs},
author = {Chris Rockwell and Justin Johnson and David F. Fouhey},
journal= {arXiv preprint arXiv:2208.08988},
year = {2023}
}
Comments
Accepted to 3DV 2022; Project Page: https://crockwell.github.io/rel_pose/ Revision: Fixed Epipolar Lines in Figure 3, Figure 10