The 8-Point Algorithm as an Inductive Bias for Relative Pose Prediction by ViTs

08/18/2022
by   Chris Rockwell, et al.
9

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.

READ FULL TEXT

page 6

page 8

page 14

page 15

page 16

page 17

page 18

page 19

research
05/19/2011

An Algorithmic Solution to the Five-Point Pose Problem Based on the Cayley Representation of Rotations

We give a new algorithmic solution to the well-known five-point relative...
research
11/27/2022

GRelPose: Generalizable End-to-End Relative Camera Pose Regression

This paper proposes a generalizable, end-to-end deep learning-based meth...
research
06/11/2023

2-D SSM: A General Spatial Layer for Visual Transformers

A central objective in computer vision is to design models with appropri...
research
12/09/2020

Positional Encoding as Spatial Inductive Bias in GANs

SinGAN shows impressive capability in learning internal patch distributi...
research
03/05/2023

Learning to Localize in Unseen Scenes with Relative Pose Regressors

Relative pose regressors (RPRs) localize a camera by estimating its rela...
research
07/27/2022

Convolutional Embedding Makes Hierarchical Vision Transformer Stronger

Vision Transformers (ViTs) have recently dominated a range of computer v...
research
06/30/2023

Act3D: Infinite Resolution Action Detection Transformer for Robotic Manipulation

3D perceptual representations are well suited for robot manipulation as ...

Please sign up or login with your details

Forgot password? Click here to reset