AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

LMT-GP: Combined Latent Mean-Teacher and Gaussian Process for Semi-supervised Low-light Image Enhancement

Ye Yu, Fengxin Chen, Jun Yu, Zhen Kan

ECCV 2024 · Label-efficient and robust learning › Semi-supervised learning · approach: Pseudo-labeling for specialized tasks · 14 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
add mean-teacher-assisted GP regression loss to basic mean-teacher (ablation)LMT-GP · LOLPSNR: 22.73 -> 24.12Improved
use constant-covariance variant of GPR loss instead of full L_gpr (ablation)LMT-GP · LOLPSNR: 23.55 -> 24.12Improved
add pseudo-label adaptation module (PAM) (ablation)LMT-GP · LOLPSNR: 21.54 -> 24.12Improved
select PAM latent vectors by PSNR instead of SSIM (ablation)PAM · LOLPSNR: 22.65 -> 24.12Improved
add perceptual loss (ablation)LMT-GP · LOLPSNR: 21.66 -> 24.12Improved
add SSIM loss (ablation)LMT-GP · LOLPSNR: 23.87 -> 24.12Improved
add unlabeled loss (ablation)LMT-GP · LOLSSIM: PSNR 23.85 -> 24.12, SSIM 0.906 -> 0.902Mixed
use LMT-GP-enhanced images as preprocessing for night segmentationHRNet on ACDC · ACDCmIoU: 31.66 -> 39.33Improved
use LMT-GP preprocessing for face detectionYOLOv5 on Darkface · DarkfacemAP (%): 40.47 -> 43.92Improved
apply EnGAN as preprocessing for face detectionYOLOv5 on Darkface · DarkfacemAP (%): 40.47 -> 37.34Hurt

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