AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Spatially-Adaptive Filter Units for Deep Neural Networks

Domen Tabernik, Matej Kristan, Aleš Leonardis

CVPR 2018 · Architectures and efficiency › CNN design · approach: Invertible and wide-network analysis · 5 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
replace regular convolution filters with learned displaced aggregation unitsAlexNet · ILSVRC 2012 validationtop-1 accuracy (%): 56.99 -> 56.89No clear effect
replace regular convolution filters with learned displaced aggregation unitsAlexNet · ILSVRC 2012 validationconvolutional-layer parameters: 3.7 mio -> 2.3 mioImproved
use learned DAU filters instead of dilated convolutionAlexNet segmentation network · PASCAL VOC 2011 validationmean IU: 45.57 -> 47.22Improved
vary Gaussian aggregation variance from 0.3^2 to 0.8^2 (ablation)DAU-ConvNet · CIFAR10classification performance: 82.9, 83.4, 83.8, 83.6, 82.9, 82.8No clear effect
increase DAUs per filter from 1 to 6 (ablation)DAU-ConvNet · CIFAR10classification performance: 82.9 -> 84.1Improved
use 1–2 DAUs per filter rather than more units (ablation)DAU-ConvNet · ILSVRC 2012 validationtop-1 accuracy (%): 57.3; 56.9; 56.4No clear effect
remove units below relative amplification threshold 0.01 (ablation)Large DAU-ConvNet · ILSVRC 2012 validationtop-1 accuracy (%): 57.3 -> 57.3No clear effect
remove units below relative amplification threshold 0.02 (ablation)Large DAU-ConvNet · ILSVRC 2012 validationtop-1 accuracy (%): 57.3 -> 57.1No clear effect
remove units below relative amplification threshold 0.05, 0.10, or 0.25 (ablation)Large DAU-ConvNet · ILSVRC 2012 validationtop-1 accuracy (%): 57.3 -> 40.1; 28.3; 0.2Hurt
remove units below relative amplification threshold 0.01 or 0.02 (ablation)Medium DAU-ConvNet · ILSVRC 2012 validationtop-1 accuracy (%): 56.9 -> 56.8; 56.7No clear effect
remove units below relative amplification threshold 0.05, 0.10, or 0.25 (ablation)Medium DAU-ConvNet · ILSVRC 2012 validationtop-1 accuracy (%): 56.9 -> 54.8; 47.4; 1.9Hurt
remove units below relative amplification threshold 0.01 or 0.02 (ablation)Small DAU-ConvNet · ILSVRC 2012 validationtop-1 accuracy (%): 56.4 -> 56.4No clear effect

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