AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Learning to Zoom: a Saliency-Based Sampling Layer for Neural Networks

Adria Recasens, Petr Kellnhofer, Simon Stent, Wojciech Matusik, Antonio Torralba

ECCV 2018 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis · 159 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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 the saliency sampler layerAlexNet · GazeCapture (iPad)distance error (cm): 5.44 -> 3.29Improved
add the saliency sampler layerAlexNet · GazeCapture (iPhone)distance error (cm): 2.63 -> 2.03Improved
add deformable convolutions instead of the samplerAlexNet · GazeCapture (iPad)distance error (cm): 5.44 -> 5.21No clear effect
add affine STNAlexNet · GazeCapture (iPad)distance error (cm): 5.44 -> 4.33Improved
replace affine STN by TPS STNAlexNet + STN · GazeCapture (iPad)distance error (cm): 4.33 -> 4.44No clear effect
estimate the sampling grid directly without a saliency mapAlexNet · GazeCapture (iPad)distance error (cm): 5.44 -> 3.91Improved
add the saliency sampler layerResNet-101 · iNaturalist 2017 valtop-1 accuracy (%): 60 -> 65Improved
add the saliency sampler layerResNet-101 · iNaturalist 2017 valtop-5 accuracy (%): 83 -> 86Improved
add CAM-based cropsResNet-101 · iNaturalist 2017 valtop-1 accuracy (%): 60 -> 62Improved
add the saliency sampler layerInception V3 · iNaturalist 2017 valtop-1 accuracy (%): 64 -> 66Improved
add the saliency sampler layerResNet-50 · CUB-200top-1 accuracy (%): 81.6 -> 84.5Improved
deepen the saliency network from none to 6 layers (ablation)RN-101 + Saliency Sampler · iNaturalist 2017 valtop-1 accuracy (%): 60 -> 62Improved

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