AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Invertible Rescaling Network and Its Extensions

Mingqing Xiao, Shuxin Zheng, Chang Liu, Zhouchen Lin, Tie‐Yan Liu

IJCV 2022 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Expressive invertible architectures · 38 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 L2 LR guidance loss instead of L1 (ablation)IRN · Set5 (4x rescaling)PSNR (dB): 34.75 -> 36.19Improved
use L1 HR reconstruction loss instead of L2 (ablation)IRN · Set5 (4x rescaling)PSNR (dB): 35.93 -> 36.19Improved
add partial distribution matching loss (ablation)IRN · Set5 (4x rescaling)PSNR (dB): 36.12 -> 36.19Improved
add partial distribution matching loss (ablation)IRN · Urban100 (4x rescaling)PSNR (dB): 30.91 -> 31.41Improved
set guidance weight lambda2 to 16 instead of 160 (ablation)IRN · Set5 (4x rescaling)PSNR (dB): 35.94 -> 36.19Improved
set guidance weight lambda2 to 16 instead of 1.6 (ablation)IRN · Set5 (4x rescaling)PSNR (dB): 35.72 -> 36.19Improved
raise guidance weight lambda2 from 16 to 160 (ablation)IRN · Set5 (4x, downscaled LR vs bicubic LR)PSNR (dB): 44.60 -> 50.14Improved
replace IRN with IRN+ (full distribution matching and perceptual loss)IRN · Set5 (4x rescaling)FID: 33.88 -> 11.15Improved
replace IRN with IRN+IRN · Set5 (4x rescaling)LPIPS: 0.078 -> 0.031Improved
replace IRN with IRN+IRN · Set5 (4x rescaling)PSNR (dB): 36.19 -> 33.59Hurt
use IRN instead of CAR and EDSRCAR & EDSR · Set5 (4x rescaling)PSNR (dB): 33.88 -> 36.19Improved
use IRN instead of CAR and EDSRCAR & EDSR · Set5 (2x rescaling)PSNR (dB): 38.94 -> 43.99Improved

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