AI Paper Atlas

Research map / Image generation / GANs

Neural Space-Filling Curves

Hanyu Wang, Kamal Gupta, Larry Davis, Abhinav Shrivastava

ECCV 2022 · Image generation › GANs · approach: Large-scale GAN latent analysis · 6 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 Neural SFC order instead of Hilbert curveLZW compression · MNISTLZW size in bytes: 182.7 (+7.3) -> 171.1 (-4.3)Improved
use Neural SFC order instead of raster scanLZW compression · Fashion-MNISTLZW size in bytes: 425.8 -> 412.4 (-13.4)Improved
use Neural SFC order instead of raster scanLZW compression · FFHQ (32x32)LZW size in bytes: 688.0 -> 678.3 (-9.7)Improved
use Neural SFC order instead of raster scanLZW compression · TGIFLZW size in bytes per frame: 563.9 -> 556.9 (-7.0)Improved
use Hilbert curve instead of raster scanLZW compression · MNISTLZW size in bytes: 175.4 -> 182.7 (+7.3)Hurt
use Neural SFC instead of Dafner SFC averaged over image setautocorrelation · MNISTlag-6 autocorrelation: 0.401 -> 0.558Improved
use Neural SFC instead of Hilbert curveautocorrelation · Fashion-MNISTlag-10 autocorrelation: 0.647 -> 0.705Improved
use Neural SFC instead of Hilbert curveautocorrelation · FFHQ (32x32)lag-6 autocorrelation: 0.924 -> 0.943Improved
average Dafner SFC weights over the image set instead of using Hilbert curveautocorrelation · MNISTlag-6 autocorrelation: 0.475 -> 0.401Hurt
train with lag k=4 and 6 together instead of k=6 only (ablation)Neural SFC · MNISTlag-k autocorrelation (k=4 to 6): even better autocorrelations from k=4 to 6Improved
train with k=4, 6, 8 instead of k=6 only (ablation)Neural SFC · MNISTlag-k autocorrelation: worse performanceHurt
train for k=6 instead of other single k values (ablation)Neural SFC · MNISTlag-k autocorrelation: k=6 generally best single kImproved

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