AI Paper Atlas

Research map / Image generation / Editing and style transfer

Tunable Convolutions With Parametric Multi-Loss Optimization

Matteo Maggioni, Thomas Tanay, Francesca Babiloni, Steven McDonagh, Aleš Leonardis

CVPR 2023 · Image generation › Editing and style transfer · approach: Image-to-image translation · 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 shared interactive parameters to interpolate tunable kernels and reconstruction/noise-preservation lossesResNet denoising · Kodak and CBSD68, intermediate parameter setting (0.50, 0.50)PSNR: 27.95 -> 28.46; 27.95 -> 28.12Improved
Use shared interactive parameters to interpolate tunable kernels and reconstruction/noise-preservation lossesResNet denoising · Kodak and CBSD68, intermediate parameter setting (0.50, 0.50)PSNRη: 34.45 -> 35.56; 34.33 -> 34.98Improved
Train tunable convolutions for joint denoising and deblurring rather than conditioning on degradation parametersJoint restoration model · Kodak, ω=(1.00, 1.00)PSNR: 20.65 -> 24.84Improved
Train tunable convolutions for joint denoising and deblurring rather than conditioning on degradation parametersJoint restoration model · Kodak, ω=(1.00, 1.00)LPIPS: 0.218 -> 0.205Improved
Use shared interactive parameters to tune reconstruction and adversarial objectivesResNet ×4 super-resolution · Kodak, ω=(0.50, 0.50)PSNR: 27.08 -> 27.12Improved
Use shared interactive parameters to tune reconstruction and adversarial objectivesResNet ×4 super-resolution · CBSD68, ω=(0.50, 0.50)LPIPS: 0.125 -> 0.115Improved
Replace fixed NAFNet convolutions with tunable convolutionsNAFNet · SIDD real raw image denoising, fidelity settingPSNR: 39.96 -> 39.90No clear effect
Replace fixed SwinIR convolutions and layers with tunable variantsSwinIR · Set5 ×4 super-resolution, fidelity settingPSNR: 32.72 -> 32.57No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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