Ctrl K

Data underlying the publication/thesis chapter: Learning to Adapt to Position Bias in Vision Transformer Classifiers

Data underlying the publication/thesis chapter: Learning to Adapt to Position Bias in Vision Transformer Classifiers

2
contributors

Description

This article proposes a novel method to measure learned position bias in Vision Transformers, and investigates which model design choices affect learned equivariance.

Logo of Data underlying the publication/thesis chapter: Learning to Adapt to Position Bias in Vision Transformer Classifiers
Keywords
computer vision
position bias
position embeddings
Vision Transformers
License
  • CC-BY-4.0
</>Source code
Packages
data.4tu.nl

Reference papers

Contributors

JvG
Jan van Gemert
RB
Robert-Jan Bruintjes

Member of community

4TU