atlas

Research map / Architectures and efficiency / CNN design

Warped Convolutions: Efficient Invariance to Spatial Transformations

ICML 2017 · Architectures and efficiency › CNN design · approach: Steerable rotation-equivariant CNNs · 23 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add log-polar warp (scale and rotation) before standard CNNCNN+softargmax on Google Earth · Google Earthrotation error (degrees): 9.36 -> 8.29Helped
add log-polar warp before CNNCNN+softargmax on Google Earth · Google Earthscale error (pixels): 4.87 -> 4.79Helped
replace soft argmax with fully-connected layer (ablation)Warped CNN architecture on Google Earth · Google Earthrotation error (degrees): 8.29 -> 22.54Hurt
add 3D sphere warp before CNN with soft argmaxface pose estimation · AFLWyaw error (degrees): 12.56 -> 7.07 vs CNN+FCHelped
add 3D sphere warp before CNN with soft argmaxface pose estimation · AFLWpitch error (degrees): 6.59 -> 5.28 vs CNN+FCHelped
Dieleman-style discrete 90-degree rotations and flips instead of continuous warpGoogle Earth · Google Earthscale error (pixels): 4.79 -> 4.29 (baseline better)Helped

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy