Skip to content

In this work we explore whether MLPs are naturally biased towards using shape or texture as the main feature for image classification decisions.

Notifications You must be signed in to change notification settings

daniele-roncaglioni/MLPs_shape_vs_texture_bias

Repository files navigation

Scaling MLPs : Shape vs Texture Bias

This project was based on the repo Scaling MLPs (see accompanying paper) by Gregor Bachmann et al.

Results

Paper Paper Paper Paper Paper

About

In this work we explore whether MLPs are naturally biased towards using shape or texture as the main feature for image classification decisions.

Topics

Resources

Stars

Watchers

Forks