AI Paper Atlas

Research map / Graph learning / GNN architectures

SelectionConv: Convolutional Neural Networks for Non-Rectilinear Image Data

David Hart, Michael Whitney, Bryan Morse

ECCV 2022 · Graph learning › GNN architectures · approach: Deep graph convolution and simplification · 4 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
use SelectionConv on a spherical graph instead of applying the FCN to equirectangular imagesFCN ResNet-50 · Stanford 2D-3D-S validationaverage IoU: 32.57 -> 36.29Improved
use SelectionConv instead of distortion-aware convolutionsFCN ResNet-50 · Stanford 2D-3D-S validationaverage IoU: 34.56 -> 36.29Improved
copy CNN weights into SelectionConv graph networkVGG-11 · CIFAR-10 validationaccuracy: 84.5 -> 84.5No clear effect
add small random spatial perturbations to graph points (ablation)SelectionConv VGG-11 · CIFAR-10 validationaccuracyNo clear effect
use SelectionConv spherical graph instead of optimization-based stylizationfeed-forward style transfer · 360-degree style transferruntime: 8-10 minutes -> 15-20 secondsImproved
process SLIC superpixel graph instead of full-resolution image on CPUmonocular depth estimator · 4K depth predictionruntime: 25.9 -> 5.1 secondsImproved

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